[opensuse] swap not available
Hi, On my new 10.3 install I set up 2 drives in raid1 mirror, using software raid in yast. All seems to be working fine except in Kinfocenter>memory swap is shown as not available. Yast shows swap with an "*" beside it. My partitions are set up as follows: primary /dev/md0 /boot extended /dev/md1 /swap /dev/md2 / /dev/md3 /home /dev/md4 /share I'm not that familiar with tweaking swap and have only set it up in yast>partitioner before with no previous problems. This is my first raid setup so there may be an issue with that, but again, all others partitions are working fine. I tried editing swap in yast to format it again as /swap but it failed with an error code -3004. Anyone know what this problem may be? Many thanks, JIm F -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
On Tuesday 04 December 2007 12:48:46 Jim Flanagan wrote:
Hi,
On my new 10.3 install I set up 2 drives in raid1 mirror, using software raid in yast. All seems to be working fine except in Kinfocenter>memory swap is shown as not available. Yast shows swap with an "*" beside it. My partitions are set up as follows: primary /dev/md0 /boot extended /dev/md1 /swap /dev/md2 / /dev/md3 /home /dev/md4 /share
I'm not that familiar with tweaking swap and have only set it up in yast>partitioner before with no previous problems. This is my first raid setup so there may be an issue with that, but again, all others partitions are working fine. I tried editing swap in yast to format it again as /swap but it failed with an error code -3004.
Anyone know what this problem may be?
Many thanks,
JIm F
I don't normally SoftRAID the swap partitions as it would be faster just to have multiple ones instead (you're not limited to one). Have you tried tryping swapon /dev/md1 to see if it starts working? Matthew -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
On Tuesday 04 December 2007 06:46:03 Matthew Stringer wrote:
On Tuesday 04 December 2007 12:48:46 Jim Flanagan wrote:
Hi,
On my new 10.3 install I set up 2 drives in raid1 mirror, using software raid in yast. All seems to be working fine except in Kinfocenter>memory swap is shown as not available. Yast shows swap with an "*" beside it. My partitions are set up as follows: primary /dev/md0 /boot extended /dev/md1 /swap /dev/md2 / /dev/md3 /home /dev/md4 /share
I'm not that familiar with tweaking swap and have only set it up in yast>partitioner before with no previous problems. This is my first raid setup so there may be an issue with that, but again, all others partitions are working fine. I tried editing swap in yast to format it again as /swap but it failed with an error code -3004.
Anyone know what this problem may be?
Many thanks,
JIm F
Try running swapon -s and see if the device is listed. Also, cat /proc/mdstat. Your swap device should show up in the swapon -s command. It should show up in mdstat but may be listed as (auto-read-only). I had a similar issue on one of my hosts and ended up having to change the boot flags. I edited /boot/grub/menu.lst and changed the item resume=/dev/md1 to noresume. This disables the ability to hibernate the machine, but allowed my swap to work properly. Hope this helps. -jc -- ******************************** J.C. Polanycia Information Technology Services University of Colorado -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Jc Polanycia wrote:
On Tuesday 04 December 2007 06:46:03 Matthew Stringer wrote:
On Tuesday 04 December 2007 12:48:46 Jim Flanagan wrote:
Hi,
On my new 10.3 install I set up 2 drives in raid1 mirror, using software raid in yast. All seems to be working fine except in Kinfocenter>memory swap is shown as not available. Yast shows swap with an "*" beside it. My partitions are set up as follows: primary /dev/md0 /boot extended /dev/md1 /swap /dev/md2 / /dev/md3 /home /dev/md4 /share
I'm not that familiar with tweaking swap and have only set it up in yast>partitioner before with no previous problems. This is my first raid setup so there may be an issue with that, but again, all others partitions are working fine. I tried editing swap in yast to format it again as /swap but it failed with an error code -3004.
Anyone know what this problem may be?
Many thanks,
JIm F
Try running swapon -s and see if the device is listed. Also, cat /proc/mdstat. Your swap device should show up in the swapon -s command. It should show up in mdstat but may be listed as (auto-read-only). I had a similar issue on one of my hosts and ended up having to change the boot flags. I edited /boot/grub/menu.lst and changed the item resume=/dev/md1 to noresume. This disables the ability to hibernate the machine, but allowed my swap to work properly.
Hope this helps.
-jc
swapon -s shows one line like this.... Filename Type Size Use Priority with nothing else listed below. cat /proc/mdstat does show md1 as active (auto-read-only) Do you mean replace "resume=/dev/md1" with "noresume"? Also, while at this point I don't envision hibernating this machine, you never know. Is there a different fix without disabling hibernate? Will subsequent grub installs pick up this noresume flag? Many thanks, Jim F -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
On Dec 4, 2007, at 10:06 AM, Jim Flanagan wrote:
error code -3004
You haven't defined any swap space it seems. Swap space is space reserved on your hard drive for the machine to use for paging out of memory. It is defined in the partition manager as a FILESYSTEM TYPE which is named swap. If you have chosen this you will never be allowed to choose a mount point for it since by definition there isn't one, the box for mount point will also be greyed out. My advice would be to delete your current swap partition (/swap) and recreate a new correct swap partition. Cheers Todd Smith Systems Administrator --------------------------------------------- Soho VFX - Visual Effects Studio 99 Atlantic Avenue, Suite 303 Toronto, Ontario, M6K 3J8 (416) 516-7863 http://www.sohovfx.com --------------------------------------------- -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
On Tue, 2007-12-04 at 09:06 -0600, Jim Flanagan wrote:
Jc Polanycia wrote:
On Tuesday 04 December 2007 06:46:03 Matthew Stringer wrote:
On Tuesday 04 December 2007 12:48:46 Jim Flanagan wrote:
Hi,
On my new 10.3 install I set up 2 drives in raid1 mirror, using software raid in yast. All seems to be working fine except in Kinfocenter>memory swap is shown as not available. Yast shows swap with an "*" beside it. My partitions are set up as follows: primary /dev/md0 /boot extended /dev/md1 /swap /dev/md2 / /dev/md3 /home /dev/md4 /share
I'm not that familiar with tweaking swap and have only set it up in yast>partitioner before with no previous problems. This is my first raid setup so there may be an issue with that, but again, all others partitions are working fine. I tried editing swap in yast to format it again as /swap but it failed with an error code -3004.
Anyone know what this problem may be?
Many thanks,
JIm F
Try running swapon -s and see if the device is listed. Also, cat /proc/mdstat. Your swap device should show up in the swapon -s command. It should show up in mdstat but may be listed as (auto-read-only). I had a similar issue on one of my hosts and ended up having to change the boot flags. I edited /boot/grub/menu.lst and changed the item resume=/dev/md1 to noresume. This disables the ability to hibernate the machine, but allowed my swap to work properly.
Hope this helps.
-jc
swapon -s shows one line like this.... Filename Type Size Use Priority
with nothing else listed below.
cat /proc/mdstat does show md1 as active (auto-read-only)
Do you mean replace "resume=/dev/md1" with "noresume"?
Also, while at this point I don't envision hibernating this machine, you never know. Is there a different fix without disabling hibernate? Will subsequent grub installs pick up this noresume flag?
Many thanks,
Jim F What does the swap partition look like in /etc/fstab?
You can type 'cat /etc/fstab | grep swap' and paste the results here. -- ---Bryen--- -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Bryen wrote:
On Tue, 2007-12-04 at 09:06 -0600, Jim Flanagan wrote:
Jc Polanycia wrote:
On Tuesday 04 December 2007 06:46:03 Matthew Stringer wrote:
On Tuesday 04 December 2007 12:48:46 Jim Flanagan wrote:
Hi,
On my new 10.3 install I set up 2 drives in raid1 mirror, using software raid in yast. All seems to be working fine except in Kinfocenter>memory swap is shown as not available. Yast shows swap with an "*" beside it. My partitions are set up as follows: primary /dev/md0 /boot extended /dev/md1 /swap /dev/md2 / /dev/md3 /home /dev/md4 /share
I'm not that familiar with tweaking swap and have only set it up in yast>partitioner before with no previous problems. This is my first raid setup so there may be an issue with that, but again, all others partitions are working fine. I tried editing swap in yast to format it again as /swap but it failed with an error code -3004.
Anyone know what this problem may be?
Many thanks,
JIm F
Try running swapon -s and see if the device is listed. Also, cat /proc/mdstat. Your swap device should show up in the swapon -s command. It should show up in mdstat but may be listed as (auto-read-only). I had a similar issue on one of my hosts and ended up having to change the boot flags. I edited /boot/grub/menu.lst and changed the item resume=/dev/md1 to noresume. This disables the ability to hibernate the machine, but allowed my swap to work properly.
Hope this helps.
-jc
swapon -s shows one line like this.... Filename Type Size Use Priority
with nothing else listed below.
cat /proc/mdstat does show md1 as active (auto-read-only)
Do you mean replace "resume=/dev/md1" with "noresume"?
Also, while at this point I don't envision hibernating this machine, you never know. Is there a different fix without disabling hibernate? Will subsequent grub installs pick up this noresume flag?
Many thanks,
Jim F
What does the swap partition look like in /etc/fstab?
You can type 'cat /etc/fstab | grep swap' and paste the results here.
That returns... /dev/md1 swap swap defaults 0 0 Thanks, Jim F -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Matthew Stringer wrote:
On Tuesday 04 December 2007 12:48:46 Jim Flanagan wrote:
Hi,
On my new 10.3 install I set up 2 drives in raid1 mirror, using software raid in yast. All seems to be working fine except in Kinfocenter>memory swap is shown as not available. Yast shows swap with an "*" beside it. My partitions are set up as follows: primary /dev/md0 /boot extended /dev/md1 /swap /dev/md2 / /dev/md3 /home /dev/md4 /share
I'm not that familiar with tweaking swap and have only set it up in yast>partitioner before with no previous problems. This is my first raid setup so there may be an issue with that, but again, all others partitions are working fine. I tried editing swap in yast to format it again as /swap but it failed with an error code -3004.
Anyone know what this problem may be?
Many thanks,
JIm F
I don't normally SoftRAID the swap partitions as it would be faster just to have multiple ones instead (you're not limited to one).
Have you tried tryping swapon /dev/md1 to see if it starts working?
Matthew
Running swapon /dev/md1 returns "invalid argument". I looked at swapon --help but am unsure which option I should choose. Jim F -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Matthew Stringer wrote:
On Tuesday 04 December 2007 12:48:46 Jim Flanagan wrote:
Hi,
On my new 10.3 install I set up 2 drives in raid1 mirror, using software raid in yast. All seems to be working fine except in Kinfocenter>memory swap is shown as not available. Yast shows swap with an "*" beside it. My partitions are set up as follows: primary /dev/md0 /boot extended /dev/md1 /swap /dev/md2 / /dev/md3 /home /dev/md4 /share
I'm not that familiar with tweaking swap and have only set it up in yast>partitioner before with no previous problems. This is my first raid setup so there may be an issue with that, but again, all others partitions are working fine. I tried editing swap in yast to format it again as /swap but it failed with an error code -3004.
Anyone know what this problem may be?
Many thanks,
JIm F
I don't normally SoftRAID the swap partitions as it would be faster just to have multiple ones instead (you're not limited to one).
Given one of the goals of RAID is to keep the system running when a drive fails, what happens when a drive containing swap croaks? -- Use OpenOffice.org http://www.openoffice.org -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
On Tuesday 04 December 2007 15:27:06 James Knott wrote:
Matthew Stringer wrote:
I don't normally SoftRAID the swap partitions as it would be faster just to have multiple ones instead (you're not limited to one).
Given one of the goals of RAID is to keep the system running when a drive fails, what happens when a drive containing swap croaks?
Your available swap space would be reduced, doesn't cause the system to fail. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
On Tuesday 04 December 2007 10:02:45 Matthew Stringer wrote:
On Tuesday 04 December 2007 15:27:06 James Knott wrote:
Matthew Stringer wrote:
I don't normally SoftRAID the swap partitions as it would be faster just to have multiple ones instead (you're not limited to one).
Given one of the goals of RAID is to keep the system running when a drive fails, what happens when a drive containing swap croaks?
Your available swap space would be reduced, doesn't cause the system to fail.
Actually it can cause the system to fail, depending on what was swapped to the device that failed. At the very least applications that had pages swapped to the failed swap device would crash. -jc -- ******************************** J.C. Polanycia Information Technology Services University of Colorado 303.492.3887 JC.Polanycia@Colorado.EDU -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Jc Polanycia wrote:
On Tuesday 04 December 2007 10:02:45 Matthew Stringer wrote:
On Tuesday 04 December 2007 15:27:06 James Knott wrote:
Matthew Stringer wrote:
I don't normally SoftRAID the swap partitions as it would be faster just to have multiple ones instead (you're not limited to one). Given one of the goals of RAID is to keep the system running when a drive fails, what happens when a drive containing swap croaks? Your available swap space would be reduced, doesn't cause the system to fail.
Actually it can cause the system to fail, depending on what was swapped to the device that failed. At the very least applications that had pages swapped to the failed swap device would crash.
They won't crash... because they can't resume execution... a process which hits a page that's swapped to a device which no longer functions will simply halt on I/O, waiting forever (literally) for the next page of code or data to get swapped back in. On the other hand, if the kernel has swapped out some data, and can't bring it back, THAT can cause a system crash. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Matthew Stringer wrote:
On Tuesday 04 December 2007 15:27:06 James Knott wrote:
Matthew Stringer wrote:
I don't normally SoftRAID the swap partitions as it would be faster just to have multiple ones instead (you're not limited to one). Given one of the goals of RAID is to keep the system running when a drive fails, what happens when a drive containing swap croaks?
Your available swap space would be reduced, doesn't cause the system to fail.
All the pages that are swapped out can no longer be swapped back in.... this can cause a kernel panic. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Matthew Stringer wrote:
On Tuesday 04 December 2007 15:27:06 James Knott wrote:
Matthew Stringer wrote:
I don't normally SoftRAID the swap partitions as it would be faster just to have multiple ones instead (you're not limited to one). Given one of the goals of RAID is to keep the system running when a drive fails, what happens when a drive containing swap croaks?
Your available swap space would be reduced, doesn't cause the system to fail.
And when it goes to retrieve the contents of that swap that's no longer there? Drives do fail occasionally, when the system is running. -- Use OpenOffice.org http://www.openoffice.org -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
On Tuesday 04 December 2007 17:22:24 James Knott wrote:
Matthew Stringer wrote:
On Tuesday 04 December 2007 15:27:06 James Knott wrote:
Matthew Stringer wrote:
I don't normally SoftRAID the swap partitions as it would be faster just to have multiple ones instead (you're not limited to one).
Given one of the goals of RAID is to keep the system running when a drive fails, what happens when a drive containing swap croaks?
Your available swap space would be reduced, doesn't cause the system to fail.
And when it goes to retrieve the contents of that swap that's no longer there? Drives do fail occasionally, when the system is running.
If you swap over multiple partitions the data is automatically striped, it usually copes OK if a drive blobs Linux is fairly stable these days when it comes to read/write errors. I think you're splitting hairs if you think that sotfRAID1 gives you enough extra stability which outweighs the reduction in performance. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Matthew Stringer wrote:
On Tuesday 04 December 2007 17:22:24 James Knott wrote:
Matthew Stringer wrote:
On Tuesday 04 December 2007 15:27:06 James Knott wrote:
Matthew Stringer wrote:
I don't normally SoftRAID the swap partitions as it would be faster just to have multiple ones instead (you're not limited to one).
Given one of the goals of RAID is to keep the system running when a drive fails, what happens when a drive containing swap croaks?
Your available swap space would be reduced, doesn't cause the system to fail.
And when it goes to retrieve the contents of that swap that's no longer there? Drives do fail occasionally, when the system is running.
If you swap over multiple partitions the data is automatically striped, it usually copes OK if a drive blobs Linux is fairly stable these days when it comes to read/write errors. I think you're splitting hairs if you think that sotfRAID1 gives you enough extra stability which outweighs the reduction in performance.
When you say striped, are you saying redundant info is stored, similar to RAID? If not, then there's no protection. If yes, how does striping differ from software RAID -- Use OpenOffice.org http://www.openoffice.org -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Matthew Stringer wrote:
On Tuesday 04 December 2007 17:22:24 James Knott wrote:
On Tuesday 04 December 2007 15:27:06 James Knott wrote:
Matthew Stringer wrote:
I don't normally SoftRAID the swap partitions as it would be faster just to have multiple ones instead (you're not limited to one). Given one of the goals of RAID is to keep the system running when a drive fails, what happens when a drive containing swap croaks? Your available swap space would be reduced, doesn't cause the system to fail. And when it goes to retrieve the contents of that swap that's no longer
Matthew Stringer wrote: there? Drives do fail occasionally, when the system is running.
If you swap over multiple partitions the data is automatically striped, it usually copes OK if a drive blobs Linux is fairly stable these days when it comes to read/write errors. I think you're splitting hairs if you think that sotfRAID1 gives you enough extra stability which outweighs the reduction in performance.
Striping does not provide any protection from device failure. I think you have it confused with MIRRORING, which does provide some safety against device failure. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Matthew Stringer wrote:
On Tuesday 04 December 2007 17:22:24 James Knott wrote:
On Tuesday 04 December 2007 15:27:06 James Knott wrote:
Matthew Stringer wrote:
I don't normally SoftRAID the swap partitions as it would be faster just to have multiple ones instead (you're not limited to one). Given one of the goals of RAID is to keep the system running when a drive fails, what happens when a drive containing swap croaks? Your available swap space would be reduced, doesn't cause the system to fail. And when it goes to retrieve the contents of that swap that's no longer
Matthew Stringer wrote: there? Drives do fail occasionally, when the system is running.
If you swap over multiple partitions the data is automatically striped, it usually copes OK if a drive blobs Linux is fairly stable these days when it comes to read/write errors. I think you're splitting hairs if you think that sotfRAID1 gives you enough extra stability which outweighs the reduction in performance.
There is no need to create the swap partitions as RAID drives. The simple solution is to use the ionice command to set the I/O priority of all swap partitions to the same value. The kernel then treats them in a manner similar to RAID 0. For performance reasons, you don't want anything to slow down swap. Bill Anderson WW7BA -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Bill Anderson wrote:
There is no need to create the swap partitions as RAID drives. The simple solution is to use the ionice command to set the I/O priority of all swap partitions to the same value. The kernel then treats them in a manner similar to RAID 0. For performance reasons, you don't want anything to slow down swap.
And what will happen, should a drive containing swap fail, while the system is running? Would you recommend that to someone when uptime is critical? -- Use OpenOffice.org http://www.openoffice.org -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
James Knott wrote:
Bill Anderson wrote:
There is no need to create the swap partitions as RAID drives. The simple solution is to use the ionice command to set the I/O priority of all swap partitions to the same value. The kernel then treats them in a manner similar to RAID 0. For performance reasons, you don't want anything to slow down swap.
And what will happen, should a drive containing swap fail, while the system is running? Would you recommend that to someone when uptime is critical?
If uptime were absolutely critical, I wouldn't be using Linux or any standard Unix for that matter. I would be have those apps running on QNX or a Stratus server. QNX is the only OS certified for "life and death" equipment, such as hospital monitoring machines. Stratus has a motto: "What if your brakes worked only 99.9999% of the time?" This is on a poster showing a car driving down a twisty, winding mountain road. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Aaron Kulkis wrote:
James Knott wrote:
Bill Anderson wrote:
There is no need to create the swap partitions as RAID drives. The simple solution is to use the ionice command to set the I/O priority of all swap partitions to the same value. The kernel then treats them in a manner similar to RAID 0. For performance reasons, you don't want anything to slow down swap.
And what will happen, should a drive containing swap fail, while the system is running? Would you recommend that to someone when uptime is critical?
If uptime were absolutely critical, I wouldn't be using Linux or any standard Unix for that matter. I would be have those apps running on QNX or a Stratus server.
QNX is the only OS certified for "life and death" equipment, such as hospital monitoring machines.
I wonder what their liability is if it fails. I'd love to see the EULA, just for grins. As for stratus, isn't it all about the hardware? Say, can't Linux run on some pretty high end hardware, come to think of it? We're about to install linux on big IBM iron here. But heck, even on a puny little dell, the uptime isn't bad: root@ashpool:~> w 17:18:08 up 965 days, 4:48, 1 user, load average: 2.38, 2.34, 2.38 USER TTY LOGIN@ IDLE JCPU PCPU WHAT root pts/0 17:18 0.00s 0.12s 0.02s w root@ashpool:~> Joe -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Aaron Kulkis wrote:
James Knott wrote:
Bill Anderson wrote:
There is no need to create the swap partitions as RAID drives. The simple solution is to use the ionice command to set the I/O priority of all swap partitions to the same value. The kernel then treats them in a manner similar to RAID 0. For performance reasons, you don't want anything to slow down swap.
And what will happen, should a drive containing swap fail, while the system is running? Would you recommend that to someone when uptime is critical?
If uptime were absolutely critical, I wouldn't be using Linux or any standard Unix for that matter. I would be have those apps running on QNX or a Stratus server.
QNX is the only OS certified for "life and death" equipment, such as hospital monitoring machines.
Regardless of the OS, the question remains the same. If you had a critical system, would you have swap on a non RAID drive? BTW, according to what I've read in several articles, Linux is being used in many critical applications. -- Use OpenOffice.org http://www.openoffice.org -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 The Tuesday 2007-12-04 at 19:38 -0700, Bill Anderson wrote:
There is no need to create the swap partitions as RAID drives. The simple solution is to use the ionice command to set the I/O priority of all swap partitions to the same value. The kernel then treats them in a manner similar to RAID 0. For performance reasons, you don't want anything to slow down swap.
For perfomance, yes, you are right. For safety, no, you are wrong. - -- Cheers, Carlos E. R. -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.4-svn0 (GNU/Linux) iD8DBQFHVqVDtTMYHG2NR9URAjEjAJ48WmxfBUN9qQvvWy2ZLyQESntAxACfWvnh OqtMbPKOiFXQ/hcd+CLVzx8= =Z32g -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Carlos E. R. wrote:
The Tuesday 2007-12-04 at 19:38 -0700, Bill Anderson wrote:
There is no need to create the swap partitions as RAID drives. The simple solution is to use the ionice command to set the I/O priority of all swap partitions to the same value. The kernel then treats them in a manner similar to RAID 0. For performance reasons, you don't want anything to slow down swap.
For perfomance, yes, you are right. For safety, no, you are wrong.
-- Cheers, Carlos E. R.
I don't see the safety issue as a major issue. The kernel avoids sending dirty pages to swap. Also, any time the application does a write, the dirty pages are sent to the buffer, and buffers aren't swapped. The kernel does not swap any kernel data structure. Having mirrored swap areas isn't going to protect buffers that kflushd hasn't sent to the disk. In this case, my opinion is that performance takes precedence. Bill Anderson WW7BA -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 The Wednesday 2007-12-05 at 06:48 -0700, Bill Anderson wrote:
For perfomance, yes, you are right. For safety, no, you are wrong.
I don't see the safety issue as a major issue. The kernel avoids sending dirty pages to swap. Also, any time the application does a write, the dirty pages are sent to the buffer, and buffers aren't swapped. The kernel does not swap any kernel data structure. Having mirrored swap areas isn't going to protect buffers that kflushd hasn't sent to the disk. In this case, my opinion is that performance takes precedence.
Performance takes precedence if the admin of that systems prefers performance. If the admin prefers reliability (or needs), then reliability takes precedence. If the machine is to run 24*7, perhaps with hot-swappable disks, then reliability takes precedence - and in that case swap *must* go on raid. The trick is that with swap on raid, if one disk goes down the system continues running till it is replaced with no impact on users or programs. If you don't, then applications will simply stop (in iowait for ever) or the kernel will panic. That means downtime! It is a major issue. This difference is documented. - -- Cheers, Carlos E. R. -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.4-svn0 (GNU/Linux) iD8DBQFHVsY6tTMYHG2NR9URAtUXAJ9ls+Ckzx/Go2E7bQ7lxeM6JbEYxgCfc/X9 m1to7R2J5YiCdT3vNX1wLDo= =i0IS -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
On Wednesday 05 December 2007 15:39:36 Carlos E. R. wrote:
The Wednesday 2007-12-05 at 06:48 -0700, Bill Anderson wrote:
For perfomance, yes, you are right. For safety, no, you are wrong.
I don't see the safety issue as a major issue. The kernel avoids sending dirty pages to swap. Also, any time the application does a write, the dirty pages are sent to the buffer, and buffers aren't swapped. The kernel does not swap any kernel data structure. Having mirrored swap areas isn't going to protect buffers that kflushd hasn't sent to the disk. In this case, my opinion is that performance takes precedence.
Performance takes precedence if the admin of that systems prefers performance. If the admin prefers reliability (or needs), then reliability takes precedence.
If the machine is to run 24*7, perhaps with hot-swappable disks, then reliability takes precedence - and in that case swap *must* go on raid.
The trick is that with swap on raid, if one disk goes down the system continues running till it is replaced with no impact on users or programs.
If you don't, then applications will simply stop (in iowait for ever) or the kernel will panic. That means downtime! It is a major issue.
This difference is documented.
-- Cheers, Carlos E. R.
If swap is a major issue you've clearly not got enough RAM ;) -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 The Wednesday 2007-12-05 at 16:32 -0000, Matthew Stringer wrote:
If swap is a major issue you've clearly not got enough RAM ;)
Swap is not a major issue; I didn't say that. What we are saying is that loss of swap when there is something swapped IS a major issue. And remember, having some swap can increase your system speed - now you go figure out why as an exercise ;-) - -- Cheers, Carlos E. R. -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.4-svn0 (GNU/Linux) iD8DBQFHVwTstTMYHG2NR9URArWwAJ4/1AB/xPGX2hcTbzY22/5P8hqgkwCglAww JS8X6qlEYEWcjT3Q3YZGeyo= =xQgT -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Carlos E. R. wrote:
If swap is a major issue you've clearly not got enough RAM ;)
Swap is not a major issue; I didn't say that. What we are saying is that loss of swap when there is something swapped IS a major issue.
what this remark may me think is this: is the swap really identical to ram (functionally, of course, ne speed involved). if so, is the ram more important (or less) than Hard drive? Is the ram loss more problematic? Is it more prone to happen? what do you do is your memory chip fails? jdd -- http://www.dodin.net -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
jdd wrote:
Carlos E. R. wrote:
If swap is a major issue you've clearly not got enough RAM ;)
Swap is not a major issue; I didn't say that. What we are saying is that loss of swap when there is something swapped IS a major issue.
what this remark may me think is this: is the swap really identical to ram (functionally, of course, ne speed involved).
if so, is the ram more important (or less) than Hard drive? Is the ram loss more problematic? Is it more prone to happen? what do you do is your memory chip fails?
More RAM is generally preferable to swap, however disk drives are mechanical devices and thus more likely to fail. If memory fails, you will have problems. Some systems, such as servers, use error correcting memory. -- Use OpenOffice.org http://www.openoffice.org -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
James Knott wrote:
however disk drives are mechanical devices and thus more likely to fail.
I have had at least as many ram failure than hard drive with swap, you are vulnerable to both together... jdd -- http://www.dodin.net -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
jdd wrote:
James Knott wrote:
however disk drives are mechanical devices and thus more likely to fail.
I have had at least as many ram failure than hard drive
with swap, you are vulnerable to both together...
jdd
The only computer I have with RAID also has error correcting memory, as many servers use. Also, back in the days when I was a computer tech, servicing mini-computers, hard drive failures were far more frequent than memory. -- Use OpenOffice.org http://www.openoffice.org -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
James Knott wrote:
jdd wrote:
James Knott wrote:
however disk drives are mechanical devices and thus more likely to fail.
I have had at least as many ram failure than hard drive
with swap, you are vulnerable to both together...
jdd
The only computer I have with RAID also has error correcting memory, as many servers use. Also, back in the days when I was a computer tech, servicing mini-computers, hard drive failures were far more frequent than memory.
Thanks for the great discussion on this! In my case I'm running a home server, that can be taken down for short periods of time when needed. I'm more interested in maintaining my setup and data, so I set up a raid1 to give me some redundancy here. That way if a disk encounters a problem between backups, I have some built in protection. I usually do a full backup every week, sometimes two, but not more than that. I don't have to have a failsafe setup here, but would like to not loose data. The reason I set up swap (and /boot, /, and /home) on raid was I was following the article about software raid on the opensuse wiki. That article indicated, and others I've read stated that in order to recover one lost disk with the other, ALL partitions on the disk had to be mirrored (not just /home for example). Is this true? I re-booted with the "noresume" option and can access swap now on the mirrored /dev/md1 so that's not an issue now, and I'm comfortable leaving swap mirrored on /dev/md1, but is that necessary or recommended? For performance sake I could make swap not raid, but what does that do to my recovery situation in the future if needed? There is a lot of discussion regarding raid, and one thing I've learned is that there are many different implementations. Raid is not raid is not raid. It may be that the recovery issue related to the full disk being mirrored may be related to bios (fake) raid, and not an issue with linux software raid. But I am still unclear on this. So to ask my remaining question more clearly, can I recover a lost disk with the good one if it contains a mix of raid and non-raid partitions on it, or does the whole disk need to be raid1 for recovery? Thanks again for all the great info! Jim F -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Thanks for the great discussion on this! In my case I'm running a home server, that can be taken down for short periods of time when needed. I'm more interested in maintaining my setup and data, so I set up a raid1 to give me some redundancy here. That way if a disk encounters a problem between backups, I have some built in protection. I usually do a full backup every week, sometimes two, but not more than that. I don't have to have a failsafe setup here, but would like to not loose data. Your thinking is correct and sound.
The reason I set up swap (and /boot, /, and /home) on raid was I was following the article about software raid on the opensuse wiki. That article indicated, and others I've read stated that in order to recover one lost disk with the other, ALL partitions on the disk had to be mirrored (not just /home for example). Is this true? No, that is not true. What is true is to recover, you need to be able to boot. Raid1 (assuming software raid1) can not be started by generic boot code in the MBR and needs to be booted via grub in the MBR, which is the Bios boot disk. If that dies, to be able to boot, you need to also have grub installed on the MBR of the second disk. After that, it can boot, load / and /home from only one disk if necessary. The only caveat is if you remove a disk (i.e remove the bad but do not replace) it will change disk naming), it could change disks mentioned in fstab). I have discovered if you change the BIOS boot order (i.e. to boot from
On 12/07/2007 07:51 AM, Jim Flanagan wrote: the good disk if the bad disk is the one with grub on the MBR), it changes the order of the disks as far as GRUB is concerned, hd0 is now the other disk. BUT, once booted with the new disk in place, you can partition and add them to the running degraded array without reboot.
I re-booted with the "noresume" option and can access swap now on the mirrored /dev/md1 so that's not an issue now, and I'm comfortable leaving swap mirrored on /dev/md1, but is that necessary or recommended?
For performance sake I could make swap not raid, but what does that do to my recovery situation in the future if needed? NA. Software raid can run (and even be built outside of Yast) with only one disk. It would work as normal even if a disk went bad (though when it failed and then got removed it would run very slow for a few minutes).
There is a lot of discussion regarding raid, and one thing I've learned is that there are many different implementations. Raid is not raid is not raid. It may be that the recovery issue related to the full disk being mirrored may be related to bios (fake) raid, and not an issue with linux software raid. But I am still unclear on this. So to ask my remaining question more clearly, can I recover a lost disk with the good one if it contains a mix of raid and non-raid partitions on it, or does the whole disk need to be raid1 for recovery? If you have data NOT mirrored on a bad disk, it is gone. All data should be mirrored. Since 8.2 IIRC it is no longer necessary to have a separate /boot to boot a raid partition. IMHO, swap is also not necessary, since that data is temporary by definition. I seldom see swap used. Everything else should be mirrored for data redundancy and
I did not, but it probably doesn't hurt. protection. HTH. -- Joe Morris Registered Linux user 231871 running openSUSE 10.3 x86_64 -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 The Thursday 2007-12-06 at 17:51 -0600, Jim Flanagan wrote:
Thanks for the great discussion on this! In my case I'm running a home server, that can be taken down for short periods of time when needed. I'm more interested in maintaining my setup and data, so I set up a raid1 to give me some redundancy here. That way if a disk encounters a problem between backups, I have some built in protection. I usually do a full backup every week, sometimes two, but not more than that. I don't have to have a failsafe setup here, but would like to not loose data.
It is sometimes safer to have two disks not mounted as a raid, but a backup, updated daily (or as needed) and then umounted. Why? Because if you erase by accident a tree, or a program goes mad, or the kernel crashes, you are not protected by a raid: both copies will go bad at the same time. Raid only protects you from hard disk failure.
The reason I set up swap (and /boot, /, and /home) on raid was I was following the article about software raid on the opensuse wiki. That article indicated, and others I've read stated that in order to recover one lost disk with the other, ALL partitions on the disk had to be mirrored (not just /home for example). Is this true?
Well, all the partitions you /want/ to recover must be mirrored, of course. But of course, you can mirror only partially a disk, and those things that are mirrored are recoverable, not the rest. For instance, a typical setup is to put the data on raid, but not the system, because it doesn't change as often and can be recreated, perhaps.
I re-booted with the "noresume" option and can access swap now on the mirrored /dev/md1 so that's not an issue now, and I'm comfortable leaving swap mirrored on /dev/md1, but is that necessary or recommended? For performance sake I could make swap not raid, but what does that do to my recovery situation in the future if needed?
The need to put swap on raid is only if you need your machine to survive running the same session as one of the disks goes bad. If you don't mind your machine crashing once that single time - which may be once after five years - then don't bother.
But I am still unclear on this. So to ask my remaining question more clearly, can I recover a lost disk with the good one if it contains a mix of raid and non-raid partitions on it, or does the whole disk need to be raid1 for recovery?
You can of course recover the disk. The data that is mirrored can be accessed and written to at the same time as the recovery to the new disk takes place. The data on non mirrored partitions will not - but you can get it from the backup. - -- Cheers, Carlos E. R. -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.4-svn0 (GNU/Linux) iD8DBQFHWJmOtTMYHG2NR9URAl2GAJ4vyywFbOvpZK7FwSunt/Kwz0ttBgCfcumx G9oP37oPeUx7WrPoVKaSApw= =a9hH -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
It is sometimes safer to have two disks not mounted as a raid, but a backup, updated daily (or as needed) and then umounted. Why? Because if you erase by accident a tree, or a program goes mad, or the kernel crashes, you are not protected by a raid: both copies will go bad at the same time. Raid only protects you from hard disk failure. I will go along with that. This is what an admin here at work suggested for my home setup. I used to have a soft RAID1 setup on SuSE 9.1 but
Carlos E. R. wrote: the admin suggested that I would get slightly better performance without the RAID and I get the advantage you mentioned. The admin suggesting using rsync and that has been working quite well for me. Damon Register -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
It is sometimes safer to have two disks not mounted as a raid, but a backup, updated daily (or as needed) and then umounted. Why? Because if you erase by accident a tree, or a program goes mad, or the kernel crashes, you are not protected by a raid: both copies will go bad at the same time. Raid only protects you from hard disk failure. I will go along with that. This is what an admin here at work suggested for my home setup. I used to have a soft RAID1 setup on SuSE 9.1 but
Carlos E. R. wrote: the admin suggested that I would get slightly better performance without the RAID and I get the advantage you mentioned. The admin suggesting using rsync and that has been working quite well for me. It isn't an either/or situation. you can have both, if your drives are big enough. Raid 1 has advantages, backup with rsync, etc. Personally I have raid 1, an external backup via rsync, and use a versioned backup via storeBackup to another raid1 partition. They work very well for
On 12/07/2007 07:31 PM, Damon Register wrote: their intended strengths. -- Joe Morris Registered Linux user 231871 running openSUSE 10.3 x86_64 -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Jim Flanagan wrote:
James Knott wrote:
jdd wrote:
James Knott wrote:
however disk drives are mechanical devices and thus more likely to fail.
I have had at least as many ram failure than hard drive
with swap, you are vulnerable to both together...
jdd
The only computer I have with RAID also has error correcting memory, as many servers use. Also, back in the days when I was a computer tech, servicing mini-computers, hard drive failures were far more frequent than memory.
Thanks for the great discussion on this! In my case I'm running a home server, that can be taken down for short periods of time when needed. I'm more interested in maintaining my setup and data, so I set up a raid1 to give me some redundancy here. That way if a disk encounters a problem between backups, I have some built in protection. I usually do a full backup every week, sometimes two, but not more than that. I don't have to have a failsafe setup here, but would like to not loose data.
The reason I set up swap (and /boot, /, and /home) on raid was I was following the article about software raid on the opensuse wiki. That article indicated, and others I've read stated that in order to recover one lost disk with the other, ALL partitions on the disk had to be mirrored (not just /home for example). Is this true?
I re-booted with the "noresume" option and can access swap now on the mirrored /dev/md1 so that's not an issue now, and I'm comfortable leaving swap mirrored on /dev/md1, but is that necessary or recommended? For performance sake I could make swap not raid, but what does that do to my recovery situation in the future if needed?
There is a lot of discussion regarding raid, and one thing I've learned is that there are many different implementations. Raid is not raid is not raid. It may be that the recovery issue related to the full disk being mirrored may be related to bios (fake) raid, and not an issue with linux software raid. But I am still unclear on this. So to ask my remaining question more clearly, can I recover a lost disk with the good one if it contains a mix of raid and non-raid partitions on it, or does the whole disk need to be raid1 for recovery?
My "server" is an IBM Netfinity box, which I bought ($150) to experiment with. It has 4 18 GB SCSI drives & 1 GB of memory. I have /boot on RAID 1 and created a RAID 5 array for everything else. I used LVM to partition that RAID array into the various partitions, such as /, /home, swap etc. -- Use OpenOffice.org http://www.openoffice.org -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
On Thu, 2007-12-06 at 17:26 -0500, James Knott wrote:
jdd wrote:
James Knott wrote:
however disk drives are mechanical devices and thus more likely to fail.
I have had at least as many ram failure than hard drive
with swap, you are vulnerable to both together...
jdd
The only computer I have with RAID also has error correcting memory, as many servers use. Also, back in the days when I was a computer tech, servicing mini-computers, hard drive failures were far more frequent than memory.
mem, mobo's and cpu's are most likely to suffer from ESD (and heath) The effects will show up sometimes at late as in several years. I found out that most shops haven;t got a faintest clue what ESD is. Drives will suffer not only from ESD and heath but also from wear, G-forces. So if you obtain your system from proper qualified supplier, that takes precautions and test the componentes before using them, or take those precautions yourself. chances that you suffer from mem problems is often smaller than HDD problems. Note, I assume you treat your components well, playing with overclocking, wrong ras/cas timing, overheated north-bridge can also cause "funny effects". Output of the powersupply has to meet more stricter demands, as mem and cpu are using still lower voltages these days. A spike of 0.1 volt did do anything some years ago, but can corrupt the content of your DDR today. hw -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Hans Witvliet wrote:
The only computer I have with RAID also has error correcting memory, as many servers use. Also, back in the days when I was a computer tech, servicing mini-computers, hard drive failures were far more frequent than memory.
mem, mobo's and cpu's are most likely to suffer from ESD (and heath) The effects will show up sometimes at late as in several years. I found out that most shops haven;t got a faintest clue what ESD is.
Drives will suffer not only from ESD and heath but also from wear, G-forces.
So if you obtain your system from proper qualified supplier, that takes precautions and test the componentes before using them, or take those precautions yourself. chances that you suffer from mem problems is often smaller than HDD problems.
Note, I assume you treat your components well, playing with overclocking, wrong ras/cas timing, overheated north-bridge can also cause "funny effects".
Output of the powersupply has to meet more stricter demands, as mem and cpu are using still lower voltages these days. A spike of 0.1 volt did do anything some years ago, but can corrupt the content of your DDR today.
hw
Back in the days when I supported mini-computers, everything was TTL logic, with ECL used in some critical areas. The CPU was two 15 inch square boards! A 200 MB disk pack drive was the size of a washing machine and required 3 phase power. Some of the systems I worked on were water cooled. Back in those days, I worked down to the microcode level, that is the instruction set within the CPU, that enabled it to run the instruction set the applications would use. -- Use OpenOffice.org http://www.openoffice.org -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
James Knott wrote:
Hans Witvliet wrote:
The only computer I have with RAID also has error correcting memory, as many servers use. Also, back in the days when I was a computer tech, servicing mini-computers, hard drive failures were far more frequent than memory.
mem, mobo's and cpu's are most likely to suffer from ESD (and heath) The effects will show up sometimes at late as in several years. I found out that most shops haven;t got a faintest clue what ESD is.
Drives will suffer not only from ESD and heath but also from wear, G-forces.
So if you obtain your system from proper qualified supplier, that takes precautions and test the componentes before using them, or take those precautions yourself. chances that you suffer from mem problems is often smaller than HDD problems.
Note, I assume you treat your components well, playing with overclocking, wrong ras/cas timing, overheated north-bridge can also cause "funny effects".
Output of the powersupply has to meet more stricter demands, as mem and cpu are using still lower voltages these days. A spike of 0.1 volt did do anything some years ago, but can corrupt the content of your DDR today.
hw
Back in the days when I supported mini-computers, everything was TTL logic, with ECL used in some critical areas. The CPU was two 15 inch square boards! A 200 MB disk pack drive was the size of a washing machine and required 3 phase power. Some of the systems I worked on were water cooled. Back in those days, I worked down to the microcode level, that is the instruction set within the CPU, that enabled it to run the instruction set the applications would use.
And some CPUs from IBM even had the capability of CHANGING instruction sets between processes. Like an IBM 370 could simulate several single-user IBM 360's and single-user IBM 704's by just switching microcode with each context switch. IBM's software from the era is nothing to rave about, but their hardware has always been top notch. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Aaron Kulkis wrote:
James Knott wrote:
Hans Witvliet wrote:
The only computer I have with RAID also has error correcting memory, as many servers use. Also, back in the days when I was a computer tech, servicing mini-computers, hard drive failures were far more frequent than memory.
mem, mobo's and cpu's are most likely to suffer from ESD (and heath) The effects will show up sometimes at late as in several years. I found out that most shops haven;t got a faintest clue what ESD is.
Drives will suffer not only from ESD and heath but also from wear, G-forces.
So if you obtain your system from proper qualified supplier, that takes precautions and test the componentes before using them, or take those precautions yourself. chances that you suffer from mem problems is often smaller than HDD problems.
Note, I assume you treat your components well, playing with overclocking, wrong ras/cas timing, overheated north-bridge can also cause "funny effects". Output of the powersupply has to meet more stricter demands, as mem and cpu are using still lower voltages these days. A spike of 0.1 volt did do anything some years ago, but can corrupt the content of your DDR today.
hw
Back in the days when I supported mini-computers, everything was TTL logic, with ECL used in some critical areas. The CPU was two 15 inch square boards! A 200 MB disk pack drive was the size of a washing machine and required 3 phase power. Some of the systems I worked on were water cooled. Back in those days, I worked down to the microcode level, that is the instruction set within the CPU, that enabled it to run the instruction set the applications would use.
And some CPUs from IBM even had the capability of CHANGING instruction sets between processes. Like an IBM 370 could simulate several single-user IBM 360's and single-user IBM 704's by just switching microcode with each context switch.
IBM's software from the era is nothing to rave about, but their hardware has always been top notch.
The Data General Eclipse line had a feature called "Writable Control Store", which could be used to add custom instructions to the CPU. The VAX 11/780 had it's microcode loaded from floppy at boot, but I don't recall if it was changeable in the same manner as the Eclipse WCS. -- Use OpenOffice.org http://www.openoffice.org -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
On Fri, 2007-12-07 at 12:59 -0500, James Knott wrote:
The Data General Eclipse line had a feature called "Writable Control Store", which could be used to add custom instructions to the CPU. The VAX 11/780 had it's microcode loaded from floppy at boot, but I don't recall if it was changeable in the same manner as the Eclipse WCS.
Floppy? It was a real huge 8" flop ... (still have them here) hw -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Hans Witvliet wrote:
On Fri, 2007-12-07 at 12:59 -0500, James Knott wrote:
The Data General Eclipse line had a feature called "Writable Control Store", which could be used to add custom instructions to the CPU. The VAX 11/780 had it's microcode loaded from floppy at boot, but I don't recall if it was changeable in the same manner as the Eclipse WCS.
Floppy? It was a real huge 8" flop ... (still have them here)
hw
Yep. There was an LSI-11 (microprocessor version of PDP-11) hidden in the cabinet, equipped with one or two 8", hard sectored drives. It was also used to connect the console terminal. As I recall, the command to use it as the VAX console was "STP" and Ctl-Z(?) to return to the LSI-11 console. Back in it's day, the VAX was considered a "super mini", a real hot system. But it only had the CPU power of a 386! BTW, that's where I first came across the "Adventure" game. :-) -- Use OpenOffice.org http://www.openoffice.org -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
On Fri, 2007-12-07 at 13:56 -0500, James Knott wrote:
Hans Witvliet wrote:
On Fri, 2007-12-07 at 12:59 -0500, James Knott wrote:
The Data General Eclipse line had a feature called "Writable Control Store", which could be used to add custom instructions to the CPU. The VAX 11/780 had it's microcode loaded from floppy at boot, but I don't recall if it was changeable in the same manner as the Eclipse WCS.
Floppy? It was a real huge 8" flop ... (still have them here)
hw
Yep. There was an LSI-11 (microprocessor version of PDP-11) hidden in the cabinet, equipped with one or two 8", hard sectored drives. It was also used to connect the console terminal. As I recall, the command to use it as the VAX console was "STP" and Ctl-Z(?) to return to the LSI-11 console. Back in it's day, the VAX was considered a "super mini", a real hot system. But it only had the CPU power of a 386!
BTW, that's where I first came across the "Adventure" game. :-)
I remember that we opened one of those floppies, and perforated the media several times, put it back in, and gave the damaged media to one of the newbies around at that time, Explaining, "well didn't the system complained about missing sectors..." -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
James Knott wrote:
Hans Witvliet wrote:
On Fri, 2007-12-07 at 12:59 -0500, James Knott wrote:
The Data General Eclipse line had a feature called "Writable Control Store", which could be used to add custom instructions to the CPU. The VAX 11/780 had it's microcode loaded from floppy at boot, but I don't recall if it was changeable in the same manner as the Eclipse WCS.
Floppy? It was a real huge 8" flop ... (still have them here)
hw
Yep. There was an LSI-11 (microprocessor version of PDP-11) hidden in the cabinet, equipped with one or two 8", hard sectored drives. It was also used to connect the console terminal. As I recall, the command to use it as the VAX console was "STP" and Ctl-Z(?) to return to the LSI-11 console. Back in it's day, the VAX was considered a "super mini", a real hot system. But it only had the CPU power of a 386!'
The old VAX had a beautiful assembly language with 13 different addressing modes (4 numerically different addressing modes were all used for a 6-bit "immediate data" addressing mode)...which, when paired with the brilliance of making the program counter and stack pointer general registers (and thus specifiable as the register to reference in all of those addressing modes) had a grand total of something like 20 effective addressing modes. GREAT design if you were writing assembly code. Unfortunately, all of those crazy addressing modes (like doubly-indirect or something like that) made for a real mess whenever a page fault occurs. Tthe CPU has to unwind the whole partially- complete instruction, and then do an interrupt to the VM's swapping routine, return from the swapper routine, and then restart the instruction... And if the instruction was something like insert a record into a doubly-linked list, or performing some sort of string manipulation (some of which are more complicated than what's in C's strings library), well, then... yeah....it's amazing the VAX guys ever got the design to function right. Initially, i thought RISC was a crazy idea, compared to the ultra-CISC VAX-11 instruction set...but then when I read about the issues with page faults... I was almost immediately convinced that yes indeed, the RISC guys did have a very, very important point, and that the Clipper chip and others weren't so crazy after all.
BTW, that's where I first came across the "Adventure" game. :-)
-- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Aaron Kulkis wrote:
Initially, i thought RISC was a crazy idea, compared to the ultra-CISC VAX-11 instruction set...but then when I read about the issues with page faults... I was almost immediately convinced that yes indeed, the RISC guys did have a very, very important point, and that the Clipper chip and others weren't so crazy after all.
I believe the Clipper chip was part of a U.S. government initiated encryption system that included a back door. http://en.wikipedia.org/wiki/Clipper_chip -- Use OpenOffice.org http://www.openoffice.org -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
James Knott wrote:
Aaron Kulkis wrote:
Initially, i thought RISC was a crazy idea, compared to the ultra-CISC VAX-11 instruction set...but then when I read about the issues with page faults... I was almost immediately convinced that yes indeed, the RISC guys did have a very, very important point, and that the Clipper chip and others weren't so crazy after all.
I believe the Clipper chip was part of a U.S. government initiated encryption system that included a back door.
Wrong Clipper -- poor wording on my part. Sun workstations originally used Motorola 680x0 CPUs, and then Sun developed their own CPU, which was called "Clipper". I believe it was the first CPU in the SPARC line of CPUs. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
On Fri, 07 Dec 2007 14:42:00 -0500
Aaron Kulkis
Initially, i thought RISC was a crazy idea, compared to the ultra-CISC VAX-11 instruction set...but then when I read about the issues with page faults... I was almost immediately convinced that yes indeed, the RISC guys did have a very, very important point, and that the Clipper chip and others weren't so crazy after all.
Initially, the Digital Alpha supported VMS, Digital Unix (Tru64 Unix),
and OpenVMS. There was a program (VEST) that could be used to translate
a VMS application and library to run on the Alpha. The Vested
application was both translated and emulated. Over time, the
application actually needed less emulation.
Floating point was an issue. The Alpha chip had native 64-bit IEEE
format (sign bit, 11 exponent bits, 53 mantissa bits - there was
1 hidden bit) floating point hardware (with full IEEE being executed in
PALcode or software). The VAX had a number of different hardware
floating point formats so that most of the native VAX floating point
formats were emulated on the Alpha.
--
Jerry Feldman
James Knott wrote:
Hans Witvliet wrote:
On Fri, 2007-12-07 at 12:59 -0500, James Knott wrote:
The Data General Eclipse line had a feature called "Writable Control Store", which could be used to add custom instructions to the CPU. The VAX 11/780 had it's microcode loaded from floppy at boot, but I don't recall if it was changeable in the same manner as the Eclipse WCS.
Floppy? It was a real huge 8" flop ... (still have them here)
hw
Yep. There was an LSI-11 (microprocessor version of PDP-11) hidden in the cabinet, equipped with one or two 8", hard sectored drives. It was also used to connect the console terminal. As I recall, the command to use it as the VAX console was "STP" and Ctl-Z(?) to return to the LSI-11 console. Back in it's day, the VAX was considered a "super mini", a real hot system. But it only had the CPU power of a 386!
BTW, that's where I first came across the "Adventure" game. :-
Several years ago I was given an Xerox floppy drive with a box of 8" floppies. Many "commercial" programs. Unfortunately I didn't get the computer that went with them. The lady that owned it had already given that to a school for challenged children to hammer on. *<[:o( -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Billie Walsh wrote:
James Knott wrote:
Hans Witvliet wrote:
On Fri, 2007-12-07 at 12:59 -0500, James Knott wrote:
The Data General Eclipse line had a feature called "Writable Control Store", which could be used to add custom instructions to the CPU. The VAX 11/780 had it's microcode loaded from floppy at boot, but I don't recall if it was changeable in the same manner as the Eclipse WCS.
Floppy? It was a real huge 8" flop ... (still have them here)
hw
Yep. There was an LSI-11 (microprocessor version of PDP-11) hidden in the cabinet, equipped with one or two 8", hard sectored drives. It was also used to connect the console terminal. As I recall, the command to use it as the VAX console was "STP" and Ctl-Z(?) to return to the LSI-11 console. Back in it's day, the VAX was considered a "super mini", a real hot system. But it only had the CPU power of a 386!
BTW, that's where I first came across the "Adventure" game. :-
Several years ago I was given an Xerox floppy drive with a box of 8" floppies. Many "commercial" programs. Unfortunately I didn't get the computer that went with them. The lady that owned it had already given that to a school for challenged children to hammer on. *<[:o(
as in, with large blunt pieces of metal on wooden sticks? -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Aaron Kulkis wrote:
Billie Walsh wrote:
James Knott wrote:
Hans Witvliet wrote:
On Fri, 2007-12-07 at 12:59 -0500, James Knott wrote:
The Data General Eclipse line had a feature called "Writable Control Store", which could be used to add custom instructions to the CPU. The VAX 11/780 had it's microcode loaded from floppy at boot, but I don't recall if it was changeable in the same manner as the Eclipse WCS.
Floppy? It was a real huge 8" flop ... (still have them here)
hw
Yep. There was an LSI-11 (microprocessor version of PDP-11) hidden in the cabinet, equipped with one or two 8", hard sectored drives. It was also used to connect the console terminal. As I recall, the command to use it as the VAX console was "STP" and Ctl-Z(?) to return to the LSI-11 console. Back in it's day, the VAX was considered a "super mini", a real hot system. But it only had the CPU power of a 386!
BTW, that's where I first came across the "Adventure" game. :-
Several years ago I was given an Xerox floppy drive with a box of 8" floppies. Many "commercial" programs. Unfortunately I didn't get the computer that went with them. The lady that owned it had already given that to a school for challenged children to hammer on. *<[:o(
as in, with large blunt pieces of metal on wooden sticks?
Very possibly. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Hans Witvliet wrote:
On Fri, 2007-12-07 at 12:59 -0500, James Knott wrote:
The Data General Eclipse line had a feature called "Writable Control Store", which could be used to add custom instructions to the CPU. The VAX 11/780 had it's microcode loaded from floppy at boot, but I don't recall if it was changeable in the same manner as the Eclipse WCS.
Floppy? It was a real huge 8" flop ... (still have them here)
Holding all of about 50k or so? I remember seeing some of those in an office supply store in Hafir-Al-Batin, Saudi Arabia back in 1992.
hw
-- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Aaron Kulkis wrote:
Hans Witvliet wrote:
On Fri, 2007-12-07 at 12:59 -0500, James Knott wrote:
The Data General Eclipse line had a feature called "Writable Control Store", which could be used to add custom instructions to the CPU. The VAX 11/780 had it's microcode loaded from floppy at boot, but I don't recall if it was changeable in the same manner as the Eclipse WCS.
Floppy? It was a real huge 8" flop ... (still have them here)
Holding all of about 50k or so?
I remember seeing some of those in an office supply store in Hafir-Al-Batin, Saudi Arabia back in 1992.
hw
IIRC, the ones in the VAX were about 128K, though some systems, with double sided, double density disks could do half a meg, though I recall some mention of 1 MB on a single 8" floppy. -- Use OpenOffice.org http://www.openoffice.org -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
On Fri, 2007-12-07 at 14:28 -0500, Aaron Kulkis wrote:
Hans Witvliet wrote:
On Fri, 2007-12-07 at 12:59 -0500, James Knott wrote:
The Data General Eclipse line had a feature called "Writable Control Store", which could be used to add custom instructions to the CPU. The VAX 11/780 had it's microcode loaded from floppy at boot, but I don't recall if it was changeable in the same manner as the Eclipse WCS.
Floppy? It was a real huge 8" flop ... (still have them here)
Holding all of about 50k or so?
I remember seeing some of those in an office supply store in Hafir-Al-Batin, Saudi Arabia back in 1992.
hw
No! A (for that time) decent 2MB (rediculous these days) I used it on flex & uniflex OS on my 6809. Still beautiful DISC cpu (decent instruction set computer ;) All orthoginal instructions on one A4 page. Still regret it that Intel won that battle. hw -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Hans Witvliet wrote:
On Fri, 2007-12-07 at 14:28 -0500, Aaron Kulkis wrote:
Hans Witvliet wrote:
On Fri, 2007-12-07 at 12:59 -0500, James Knott wrote:
The Data General Eclipse line had a feature called "Writable Control Store", which could be used to add custom instructions to the CPU. The VAX 11/780 had it's microcode loaded from floppy at boot, but I don't recall if it was changeable in the same manner as the Eclipse WCS.
Floppy? It was a real huge 8" flop ... (still have them here)
Holding all of about 50k or so?
I remember seeing some of those in an office supply store in Hafir-Al-Batin, Saudi Arabia back in 1992.
hw
No! A (for that time) decent 2MB (rediculous these days) I used it on flex & uniflex OS on my 6809. Still beautiful DISC cpu (decent instruction set computer ;) All orthoginal instructions on one A4 page. Still regret it that Intel won that battle.
As someone who loves assembly language, me too. However, by the time of the 80386, the Intel chips were (and remain) strongly influenced by the IBM 370 programming architecture(*) and instruction set, which was well-proven to be both efficient, and well-suited to multi-process computing (although I still think it's completely ugly!) (*) although the internal architecture has been radically changed several times, the programmer's architecture has remained fairly stable, and 100% backwards compatible.
hw
-- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
James Knott wrote:
Aaron Kulkis wrote:
James Knott wrote:
Hans Witvliet wrote:
The only computer I have with RAID also has error correcting memory, as many servers use. Also, back in the days when I was a computer tech, servicing mini-computers, hard drive failures were far more frequent than memory.
mem, mobo's and cpu's are most likely to suffer from ESD (and heath) The effects will show up sometimes at late as in several years. I found out that most shops haven;t got a faintest clue what ESD is.
Drives will suffer not only from ESD and heath but also from wear, G-forces.
So if you obtain your system from proper qualified supplier, that takes precautions and test the componentes before using them, or take those precautions yourself. chances that you suffer from mem problems is often smaller than HDD problems.
Note, I assume you treat your components well, playing with overclocking, wrong ras/cas timing, overheated north-bridge can also cause "funny effects". Output of the powersupply has to meet more stricter demands, as mem and cpu are using still lower voltages these days. A spike of 0.1 volt did do anything some years ago, but can corrupt the content of your DDR today.
hw
Back in the days when I supported mini-computers, everything was TTL logic, with ECL used in some critical areas. The CPU was two 15 inch square boards! A 200 MB disk pack drive was the size of a washing machine and required 3 phase power. Some of the systems I worked on were water cooled. Back in those days, I worked down to the microcode level, that is the instruction set within the CPU, that enabled it to run the instruction set the applications would use.
And some CPUs from IBM even had the capability of CHANGING instruction sets between processes. Like an IBM 370 could simulate several single-user IBM 360's and single-user IBM 704's by just switching microcode with each context switch.
IBM's software from the era is nothing to rave about, but their hardware has always been top notch.
The Data General Eclipse line had a feature called "Writable Control Store", which could be used to add custom instructions to the CPU. The VAX 11/780 had it's microcode loaded from floppy at boot, but I don't recall if it was changeable in the same manner as the Eclipse WCS.
I remember coming across the "Writable Control Store" when reading something, and the first thing that came to my mind was, "Boy would that create utter chaos in even a multi-programming environment, let along a multi-user timesharing environment." -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Aaron Kulkis wrote:
James Knott wrote:
The Data General Eclipse line had a feature called "Writable Control Store", which could be used to add custom instructions to the CPU. The VAX 11/780 had it's microcode loaded from floppy at boot, but I don't recall if it was changeable in the same manner as the Eclipse WCS.
I remember coming across the "Writable Control Store" when reading something, and the first thing that came to my mind was, "Boy would that create utter chaos in even a multi-programming environment, let along a multi-user timesharing environment."
The systems we had didn't come with that option and so I don't know the details of using it. However, I can well imagine a need for it in specialized applications and with proper controls on who can change it and when, it shouldn't be an issue in those environments. -- Use OpenOffice.org http://www.openoffice.org -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
On Fri, 2007-12-07 at 12:59 -0500, James Knott wrote:
Aaron Kulkis wrote:
James Knott wrote:
Back in the days when I supported mini-computers, everything was TTL logic, with ECL used in some critical areas. The CPU was two 15 inch square boards! A 200 MB disk pack drive was the size of a washing
I wrote a driver for a washing machine-sized disk, but it was only 8 MB. The older machines we used had 6' tall drums for backing store.
machine and required 3 phase power. Some of the systems I worked on were water cooled. Back in those days, I worked down to the microcode level, that is the instruction set within the CPU, that enabled it to run the instruction set the applications would use.
And some CPUs from IBM even had the capability of CHANGING instruction sets between processes. Like an IBM 370 could simulate several single-user IBM 360's and single-user IBM 704's by just switching microcode with each context switch.
Yes the driver was for a machine called VCS - variable microcode system. It had COBOL microcode, FORTRAN microcode etc. It was a development of Iliffe's basic language machine. It had a variable pot to set the clock speed. Early in the morning it was cranky and had to be booted at low speed and then you wound the clock up as it warmed up. Cheers, Dave -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 The Wednesday 2007-12-05 at 22:35 +0100, jdd wrote:
Carlos E. R. wrote:
If swap is a major issue you've clearly not got enough RAM ;)
Swap is not a major issue; I didn't say that. What we are saying is that loss of swap when there is something swapped IS a major issue.
what this remark may me think is this: is the swap really identical to ram (functionally, of course, ne speed involved).
Mmm...!
if so, is the ram more important (or less) than Hard drive? Is the ram loss more problematic? Is it more prone to happen? what do you do is your memory chip fails?
Ram loss will probably panic the kernel instantly. Worse, not only it can corrupt programs and data in memory, it can also corrupt disk data which is buffered in memory and written to disk... (it is worse for filesystems like xfs). So, the traditional response to a ram failure has been to panic and halt the computer immediately. I'm not sure what linux does, though. Typically, PCs ram was 9 bit wide: 8 bit for data, one for parity. A parity error, crash! I think there was (is?) an IRQ dedicated to this condition. Some memory modules did not have this extra bit and were thus cheaper. But you can use more bits per byte and use error correction codes to recover 1 bit error (or more) and continue working. Also, the cpu could map out a section that's known to be bad or failing. - -- Cheers, Carlos E. R. -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.4-svn0 (GNU/Linux) iD8DBQFHVz9QtTMYHG2NR9URAocMAJsF7te3Rz3c+57a7FO6pfHrDTpVOwCfRZ7v JtoOtsUtv29b+kr017UTz3s= =DfJh -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
On Wed, 2007-12-05 at 16:32 +0000, Matthew Stringer wrote:
If swap is a major issue you've clearly not got enough RAM ;)
If you (accidentily, or just once a week) need more memory then actually is in your system, swap will take care of it in a gentle way. Most of the time you will not even notice it. Without swap, OOM (Out-Of-Memory) wil give you a rude awakening. Just adding more mem (instead of some swap) is a waste of resources, or making your services/products too expensive. You are correct, but only in case the system is constantly use a lot of swap. SWAP usage is something to be minitored constantly with tools like cacti/nagios/opennms. If you use it too much and too often you have to buy more mem, or re-adjust XEN-parameters for your DOM-U HW -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
James Knott wrote:
Matthew Stringer wrote:
On Tuesday 04 December 2007 12:48:46 Jim Flanagan wrote:
Hi,
On my new 10.3 install I set up 2 drives in raid1 mirror, using software raid in yast. All seems to be working fine except in Kinfocenter>memory swap is shown as not available. Yast shows swap with an "*" beside it. My partitions are set up as follows: primary /dev/md0 /boot extended /dev/md1 /swap /dev/md2 / /dev/md3 /home /dev/md4 /share
I'm not that familiar with tweaking swap and have only set it up in yast>partitioner before with no previous problems. This is my first raid setup so there may be an issue with that, but again, all others partitions are working fine. I tried editing swap in yast to format it again as /swap but it failed with an error code -3004.
Anyone know what this problem may be?
Many thanks,
JIm F
I don't normally SoftRAID the swap partitions as it would be faster just to have multiple ones instead (you're not limited to one).
Given one of the goals of RAID is to keep the system running when a drive fails, what happens when a drive containing swap croaks?
my guess is (for unix or linux) -- kernel panic -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Jim Flanagan wrote:
I tried editing swap in yast to format it again as /swap but it failed with an error code -3004. If i understand it well mount point in partitioner for partitions formatted as swap should be "swap" not "/swap".
Kind regards Philippe -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Philippe Landau wrote:
Jim Flanagan wrote:
I tried editing swap in yast to format it again as /swap but it failed with an error code -3004. If i understand it well mount point in partitioner for partitions formatted as swap should be "swap" not "/swap".
Kind regards Philippe
there is no mount point for swap, swap is never mounted, it doesn't contain any file system jdd -- http://www.dodin.net -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
jdd wrote:
Jim Flanagan wrote:
I tried editing swap in yast to format it again as /swap but it failed with an error code -3004. If i understand it well mount point in partitioner for partitions formatted as swap should be "swap" not "/swap".
Philippe Landau wrote: there is no mount point for swap, swap is never mounted, it doesn't contain any file system Yast partitioner explicitly lists swap partition mount point as "swap", Jim your choosing "/swap" as mount point could be a source of problems.
Kind regards Philippe -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Philippe Landau wrote:
jdd wrote:
Philippe Landau wrote:
Jim Flanagan wrote:
I tried editing swap in yast to format it again as /swap but it failed with an error code -3004.
If i understand it well mount point in partitioner for partitions formatted as swap should be "swap" not "/swap".
there is no mount point for swap, swap is never mounted, it doesn't contain any file system
Yast partitioner explicitly lists swap partition mount point as "swap", Jim your choosing "/swap" as mount point could be a source of problems.
Kind regards Philippe
Sorry, my mistake, swap is not mounted as /swap, but merely swap. Thanks, Jim F -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Jim Flanagan wrote:
Hi,
On my new 10.3 install I set up 2 drives in raid1 mirror, using software raid in yast. All seems to be working fine except in Kinfocenter>memory swap is shown as not available. Yast shows swap with an "*" beside it. My partitions are set up as follows: primary /dev/md0 /boot extended /dev/md1 /swap /dev/md2 / /dev/md3 /home /dev/md4 /share
I'm not that familiar with tweaking swap and have only set it up in yast>partitioner before with no previous problems. This is my first raid setup so there may be an issue with that, but again, all others partitions are working fine. I tried editing swap in yast to format it again as /swap but it failed with an error code -3004.
Why are you mirroring swap? It offers no performance advantages, nor any significant data-protection advantages, unless the data being processed is both SUPER-valuable and very-ephemerial (such as data being collected and processed in real time from sensors in a non-repeatable or expensive-to-repeat experiment, or say, you're processing stock-market feeds, in which case 15 minute downtime = $100,000 fine, and significantly more for each additional 15 minutes of downtime. Other than something like that, or life-and-death situations (in which case, you should be using QNX), mirroring swap is both a needless waste of disk space AND also hurts your system's performance for very very little benefit. If you can, I would split that swap mirror into two separate swap partitions, and use both of them independently.
Anyone know what this problem may be?
Many thanks,
JIm F
-- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Aaron Kulkis wrote:
Jim Flanagan wrote:
Hi,
On my new 10.3 install I set up 2 drives in raid1 mirror, using software raid in yast. All seems to be working fine except in Kinfocenter>memory swap is shown as not available. Yast shows swap with an "*" beside it. My partitions are set up as follows: primary /dev/md0 /boot extended /dev/md1 /swap /dev/md2 / /dev/md3 /home /dev/md4 /share
I'm not that familiar with tweaking swap and have only set it up in yast>partitioner before with no previous problems. This is my first raid setup so there may be an issue with that, but again, all others partitions are working fine. I tried editing swap in yast to format it again as /swap but it failed with an error code -3004.
Why are you mirroring swap? It offers no performance advantages, nor any significant data-protection advantages, unless the data being processed is both SUPER-valuable and very-ephemerial (such as data being collected and processed in real time from sensors in a non-repeatable or expensive-to-repeat experiment, or say, you're processing stock-market feeds, in which case 15 minute downtime = $100,000 fine, and significantly more for each additional 15 minutes of downtime.
Other than something like that, or life-and-death situations (in which case, you should be using QNX), mirroring swap is both a needless waste of disk space AND also hurts your system's performance for very very little benefit.
If you can, I would split that swap mirror into two separate swap partitions, and use both of them independently.
Ok, I understand I could split the swap mirror into two different un-mirrored swap partitions, but still am unclear about raid1. I understood that in a mirror setup both disks need to be exactly the same, or at least the partitions on them set up the same, to be mirrored. Are you saying that only certain partitions (or only one) can be mirrored, and other partitions not mirrored on the same disk? What happens when one disk fails, with one partition mirrored and another not mirrored? I was under the belief that would toast both disks. I remember reading a discussion about /boot being under raid1 or not, and that if it was not in raid1 and other partitions were, one drive would prevent the the second one being re-mirrored after a failure. Or at least that's what I thought I read. Sorry if my questions are confusing, but I am confused. Jim F -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Am Dienstag, 4. Dezember 2007 13:48:46 schrieb Jim Flanagan:
On my new 10.3 install I set up 2 drives in raid1 mirror, using software raid in yast. All seems to be working fine except in Kinfocenter>memory swap is shown as not available. Yast shows swap with an "*" beside it. My partitions are set up as follows: primary /dev/md0 /boot extended /dev/md1 /swap /dev/md2 / /dev/md3 /home /dev/md4 /share
I'm not that familiar with tweaking swap and have only set it up in yast>partitioner before with no previous problems. This is my first raid setup so there may be an issue with that, but again, all others partitions are working fine. I tried editing swap in yast to format it again as /swap but it failed with an error code -3004.
Anyone know what this problem may be?
The reason is that suspend/resume from a raid swap partition does not work. You have three options: 1. Do not use a swap partion on raid. This ok, if a crash and a reboot after a failing disk is no problem for you. May be acceptable for Notebooks and desktop systems 2. Remove the resume=/dev/md1 from your kernel option line. Suspend/Resume from does not work. Should be fine for servers. 3. Have two swap partions one without raid for the resume and one with raid for swapping, large enough, that the first one is not really used for swapping (I haven tested this). Cheers Herbert -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Herbert Graeber wrote:
Am Dienstag, 4. Dezember 2007 13:48:46 schrieb Jim Flanagan:
On my new 10.3 install I set up 2 drives in raid1 mirror, using software raid in yast. All seems to be working fine except in Kinfocenter>memory swap is shown as not available. Yast shows swap with an "*" beside it. My partitions are set up as follows: primary /dev/md0 /boot extended /dev/md1 /swap /dev/md2 / /dev/md3 /home /dev/md4 /share
I'm not that familiar with tweaking swap and have only set it up in yast>partitioner before with no previous problems. This is my first raid setup so there may be an issue with that, but again, all others partitions are working fine. I tried editing swap in yast to format it again as /swap but it failed with an error code -3004.
Anyone know what this problem may be?
The reason is that suspend/resume from a raid swap partition does not work. You have three options:
1. Do not use a swap partion on raid. This ok, if a crash and a reboot after a failing disk is no problem for you. May be acceptable for Notebooks and desktop systems
2. Remove the resume=/dev/md1 from your kernel option line. Suspend/Resume from does not work. Should be fine for servers.
3. Have two swap partions one without raid for the resume and one with raid for swapping, large enough, that the first one is not really used for swapping (I haven tested this).
Cheers Herbert
Most intersting. Option 3 does both un-mirrored and mirrored. I'm not sure I want to test that one out, but sound plausible. I guess I'm good with putting swap on an un-mirrored partition, as long as it doesn't break the mirrors in the event of one disk failure, (see my post a few minutes ago on this). Seems un-mirrored would be faster anyway, but I want to keep the integrity of the mirrored partitions. Jim F -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 The Tuesday 2007-12-04 at 22:05 -0600, Jim Flanagan wrote:
Herbert Graeber wrote:
My partitions are set up as follows: primary /dev/md0 /boot extended /dev/md1 /swap /dev/md2 / /dev/md3 /home /dev/md4 /share
3. Have two swap partions one without raid for the resume and one with raid for swapping, large enough, that the first one is not really used for swapping (I haven tested this).
Most intersting. Option 3 does both un-mirrored and mirrored. I'm not sure I want to test that one out, but sound plausible. I guess I'm good with putting swap on an un-mirrored partition, as long as it doesn't break the mirrors in the event of one disk failure, (see my post a few minutes ago on this). Seems un-mirrored would be faster anyway, but I want to keep the integrity of the mirrored partitions.
For safety, the un-mirrored would have to be defined with lower priority, so that the mirrored swap gets filled in first. For the other part of your question, notice that you have five mirrors, and that if one disk fails you have to partition and intialize five mirrors. An alternative I haven't tested is to create one large raid, and partition the resulting raid. Then the rebuild procedure would be to re-create that single raid partition. - -- Cheers, Carlos E. R. -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.4-svn0 (GNU/Linux) iD8DBQFHVqdYtTMYHG2NR9URAmVUAJwPvXiNiTegPhp6n8wvTsS1K9YuNgCdHnQS zSaCPph4UYJ4ydvNlXs7Pm8= =AoPv -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Carlos E. R. schrieb:
The Tuesday 2007-12-04 at 22:05 -0600, Jim Flanagan wrote:
Herbert Graeber wrote:
My partitions are set up as follows: primary /dev/md0 /boot extended /dev/md1 /swap /dev/md2 / /dev/md3 /home /dev/md4 /share
3. Have two swap partions one without raid for the resume and one with raid for swapping, large enough, that the first one is not really used for swapping (I haven tested this).
Most intersting. Option 3 does both un-mirrored and mirrored. I'm not sure I want to test that one out, but sound plausible. I guess I'm good with putting swap on an un-mirrored partition, as long as it doesn't break the mirrors in the event of one disk failure, (see my post a few minutes ago on this). Seems un-mirrored would be faster anyway, but I want to keep the integrity of the mirrored partitions.
For safety, the un-mirrored would have to be defined with lower priority, so that the mirrored swap gets filled in first.
For the other part of your question, notice that you have five mirrors, and that if one disk fails you have to partition and intialize five mirrors.
An alternative I haven't tested is to create one large raid, and partition the resulting raid. Then the rebuild procedure would be to re-create that single raid partition.
It is sufficient to partition one disk. It is then an easy job to copy the partition table to other disks. Eg.: To copy the partition table from /dev/sda to /dev/sdb you can use sfdisk -d /dev/sda | sfdisk /dev/sdb Cheers, Herbert -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
participants (19)
-
Aaron Kulkis -
Bill Anderson -
Billie Walsh -
Bryen -
Carlos E. R. -
Damon Register -
Dave Howorth -
Hans Witvliet -
Herbert Graeber -
James Knott -
Jc Polanycia -
jdd -
Jerry Feldman -
Jim Flanagan -
Joe Morris (NTM) -
Joe Sloan -
M. Todd Smith -
Matthew Stringer -
Philippe Landau