[opensuse-factory] fstab, systemd and UUID

newer
[opensuse-factory] Tumbleweed -...

older
[opensuse-factory] sddm greeter...

Oliver Neukum

30 Mar 2015 30 Mar '15

07:46

Hi, I recently experienced a filesystem issue that forced me to use the rescue system and I saw two issues regarding mounting filesystems that seemed bad to them 1) swap space is mounted by UUID That is problematic. If your swap space is corrupted you use mkswap. There practically is no other option. Doing so, however, you change the UUID and your system comes up without swap pretty much without warning. That is really nasty if you need to do S4 later on. It seems to me that swap should be mounted by device name. 2) systemd while coming up will tell you that a certain UUID is unavailable and fails to boot. That message is more less useless. You pretty much won't find the partition by UUID after you reboot into the rescue system because unless you have superhuman memory or take a photograph you will not have the UUID any more. And even if you had it, you won't find it for the same reason systemd failed to find it. A useful message would tell me where the filesystem was to be mounted at. Regards Oliver -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Show replies by thread

Andrei Borzenkov

30 Mar 30 Mar

08:08

On Mon, Mar 30, 2015 at 10:46 AM, Oliver Neukum wrote:

...

Hi,

I recently experienced a filesystem issue that forced me to use the rescue system and I saw two issues regarding mounting filesystems that seemed bad to them

1) swap space is mounted by UUID That is problematic. If your swap space is corrupted you use mkswap. There practically is no other option. Doing so, however, you change the UUID and your system comes up without swap pretty much without warning.

You can change it, right? Whatever default you chose someone will be unhappy with it.

...

That is really nasty if you need to do S4 later on. It seems to me that swap should be mounted by device name.

2) systemd while coming up will tell you that a certain UUID is unavailable and fails to boot.

That message is more less useless. You pretty much won't find the partition by UUID after you reboot into the rescue system because unless you have superhuman memory or take a photograph you will not have the UUID any more. And even if you had it, you won't find it for the same reason systemd failed to find it.

A useful message would tell me where the filesystem was to be mounted at.

It should say that mount unit failed because of missing prerequisite. Mount point is both encoded in mount unit name and can be looked up in mount unit properties. -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Hannes Reinecke

08:09

New subject: [opensuse-factory] Re: fstab, systemd and UUID

On 03/30/2015 09:46 AM, Oliver Neukum wrote:

...

Hi,

I recently experienced a filesystem issue that forced me to use the rescue system and I saw two issues regarding mounting filesystems that seemed bad to them

1) swap space is mounted by UUID That is problematic. If your swap space is corrupted you use mkswap. There practically is no other option. Doing so, however, you change the UUID and your system comes up without swap pretty much without warning.

That is really nasty if you need to do S4 later on. It seems to me that swap should be mounted by device name.

That's something I've complained about, too. But the distribution policy is to use UUID :-( We should be opening a FATE request to revert that for swap partitions.

...

2) systemd while coming up will tell you that a certain UUID is unavailable and fails to boot.

That message is more less useless. You pretty much won't find the partition by UUID after you reboot into the rescue system because unless you have superhuman memory or take a photograph you will not have the UUID any more. And even if you had it, you won't find it for the same reason systemd failed to find it.

A useful message would tell me where the filesystem was to be mounted at.

Maybe a bugzilla is in order here. Cheers, Hannes -- Dr. Hannes Reinecke zSeries & Storage hare@suse.de +49 911 74053 688 SUSE LINUX GmbH, Maxfeldstr. 5, 90409 Nürnberg GF: F. Imendörffer, J. Smithard, J. Guild, D. Upmanyu, G. Norton HRB 21284 (AG Nürnberg) -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Hannes Reinecke

08:11

On 03/30/2015 10:08 AM, Andrei Borzenkov wrote:

...

On Mon, Mar 30, 2015 at 10:46 AM, Oliver Neukum wrote:

...
Hi,

I recently experienced a filesystem issue that forced me to use the rescue system and I saw two issues regarding mounting filesystems that seemed bad to them

1) swap space is mounted by UUID That is problematic. If your swap space is corrupted you use mkswap. There practically is no other option. Doing so, however, you change the UUID and your system comes up without swap pretty much without warning.

You can change it, right? Whatever default you chose someone will be unhappy with it.

Oh, grand. Offloading all decisions to the user is not a choice I'd prefer. We should be selecting sensible defaults for the installation. And IMO using 'per-UUID' mounts for swap does not fall into that category. Cheers, Hannes -- Dr. Hannes Reinecke zSeries & Storage hare@suse.de +49 911 74053 688 SUSE LINUX GmbH, Maxfeldstr. 5, 90409 Nürnberg GF: F. Imendörffer, J. Smithard, J. Guild, D. Upmanyu, G. Norton HRB 21284 (AG Nürnberg) -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Oliver Neukum

08:29

On Mon, 2015-03-30 at 11:08 +0300, Andrei Borzenkov wrote:

...

...
That message is more less useless. You pretty much won't find the partition by UUID after you reboot into the rescue system because unless you have superhuman memory or take a photograph you will not have the UUID any more. And even if you had it, you won't find it for the same reason systemd failed to find it.

A useful message would tell me where the filesystem was to be mounted at.

It should say that mount unit failed because of missing prerequisite. Mount point is both encoded in mount unit name and can be looked up in mount unit properties.

Without booting? And that's the problem. This messages are shown if the system cannot boot. They must be readable for a human being without aid of the system. Regards Oliver -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Felix Miata

08:57

Oliver Neukum composed on 2015-03-30 09:46 (UTC+0200):

...

1) swap space is mounted by UUID That is problematic.

It's a minor problem. Swap isn't needed immediately on boot, if it ever is at all. Correction is readily doable.

...

If your swap space is corrupted you use mkswap.

If you have access to mkswap, you almost certainly have access to the FSTAB that uses the old UUID, and blkid, so it's not difficult to fix. To make it easy on yourself, use mkswap -L with your choice of memorable label, then mount swap by label instead of UUID. UUID is how it got configured, not how it must stay. By label is how most of my native partitions are mounted, at the outset, and beyond. YaST2 offers that option, though a bit tedious in that it must be selected each individual mount even when selecting by-label as default before configuring any partitions individually. -- "The wise are known for their understanding, and pleasant words are persuasive." Proverbs 16:21 (New Living Translation) Team OS/2 ** Reg. Linux User #211409 ** a11y rocks! Felix Miata *** http://fm.no-ip.com/ -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Jan Engelhardt

08:59

New subject: [opensuse-factory] Re: fstab, systemd and UUID

On Monday 2015-03-30 10:09, Hannes Reinecke wrote:

...

On 03/30/2015 09:46 AM, Oliver Neukum wrote:

...
That's something I've complained about, too. But the distribution policy is to use UUID :-(

We should be opening a FATE request to revert that for swap partitions.

We should be using PARTUUID indeed. But then again, is there not some magic which just automatically goes through all swap devices, irrespective of any device or symlink name, and goes to activate them? -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Hannes Reinecke

09:17

New subject: [opensuse-factory] Re: fstab, systemd and UUID

On 03/30/2015 10:59 AM, Jan Engelhardt wrote:

...

On Monday 2015-03-30 10:09, Hannes Reinecke wrote:

...
On 03/30/2015 09:46 AM, Oliver Neukum wrote:

...
That's something I've complained about, too. But the distribution policy is to use UUID :-(

We should be opening a FATE request to revert that for swap partitions.

We should be using PARTUUID indeed. But then again, is there not some magic which just automatically goes through all swap devices, irrespective of any device or symlink name, and goes to activate them?

Hmm. PARTUUID doesn't help you for msdos partitions, right? Not everything's GPT... Cheers, Hannes -- Dr. Hannes Reinecke zSeries & Storage hare@suse.de +49 911 74053 688 SUSE LINUX GmbH, Maxfeldstr. 5, 90409 Nürnberg GF: F. Imendörffer, J. Smithard, J. Guild, D. Upmanyu, G. Norton HRB 21284 (AG Nürnberg) -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Andrei Borzenkov

09:18

On Mon, Mar 30, 2015 at 11:29 AM, Oliver Neukum wrote:

...

On Mon, 2015-03-30 at 11:08 +0300, Andrei Borzenkov wrote:

...
...
That message is more less useless. You pretty much won't find the partition by UUID after you reboot into the rescue system because unless you have superhuman memory or take a photograph you will not have the UUID any more. And even if you had it, you won't find it for the same reason systemd failed to find it.

A useful message would tell me where the filesystem was to be mounted at.

It should say that mount unit failed because of missing prerequisite. Mount point is both encoded in mount unit name and can be looked up in mount unit properties.

Without booting?

In rescue mode.

...

And that's the problem. This messages are shown if the system cannot boot. They must be readable for a human being without aid of the system.

Usually this was printed just before password prompt to enter rescue mode. -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Oliver Neukum

09:28

On Mon, 2015-03-30 at 12:18 +0300, Andrei Borzenkov wrote:

...

On Mon, Mar 30, 2015 at 11:29 AM, Oliver Neukum wrote:

...
On Mon, 2015-03-30 at 11:08 +0300, Andrei Borzenkov wrote:

...

...
...
It should say that mount unit failed because of missing prerequisite. Mount point is both encoded in mount unit name and can be looked up in mount unit properties.

Without booting?

In rescue mode.

The same message is used if booting stone cold fails totally. Regards Oliver -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Marcos Mello

10:09

New subject: [opensuse-factory] Re: fstab, systemd and UUID

Hannes Reinecke writes:

...

On 03/30/2015 10:59 AM, Jan Engelhardt wrote:

...
On Monday 2015-03-30 10:09, Hannes Reinecke wrote:

...
On 03/30/2015 09:46 AM, Oliver Neukum wrote:

...
That's something I've complained about, too. But the distribution policy is to use UUID

We should be opening a FATE request to revert that for swap partitions.

We should be using PARTUUID indeed. But then again, is there not some magic which just automatically goes through all swap devices, irrespective of any device or symlink name, and goes to activate them?

Hmm. PARTUUID doesn't help you for msdos partitions, right? Not everything's GPT...

kernel 3.7: http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=d3... util-linux 2.24: https://git.kernel.org/cgit/utils/util-linux/util-linux.git/commit/?id=d67cc... -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Marcos Mello

10:15

New subject: [opensuse-factory] Re: fstab, systemd and UUID

Oliver Neukum writes:

...

On Mon, 2015-03-30 at 12:18 +0300, Andrei Borzenkov wrote:

...
In rescue mode.

The same message is used if booting stone cold fails totally.

https://bugzilla.opensuse.org/show_bug.cgi?id=906716 ? -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Oliver Neukum

10:30

New subject: [opensuse-factory] Re: fstab, systemd and UUID

On Mon, 2015-03-30 at 10:15 +0000, Marcos Mello wrote:

...

Oliver Neukum writes:

...
On Mon, 2015-03-30 at 12:18 +0300, Andrei Borzenkov wrote:

...
In rescue mode.

The same message is used if booting stone cold fails totally.

https://bugzilla.opensuse.org/show_bug.cgi?id=906716

I am afraid not. That is specific to swap partitions and we do use "resume=". So no joy. Regards Oliver -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Olaf Hering

13:01

New subject: [opensuse-factory] Re: fstab, systemd and UUID

On Mon, Mar 30, Hannes Reinecke wrote:

...

We should be opening a FATE request to revert that for swap partitions.

Not revert, that would bring it back to by-id. Just for swap by-label would be ok, with an obvious label like swap_<kernname>. Olaf -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Jan Engelhardt

2 Apr 2 Apr

10:30

New subject: [opensuse-factory] Re: fstab, systemd and UUID

On Monday 2015-03-30 12:09, Marcos Mello wrote:

...

...
...
...
But the distribution policy is to use UUID We should be opening a FATE request to revert that for swap partitions.

We should be using PARTUUID indeed. But then again, is there not some magic which just automatically goes through all swap devices, irrespective of any device or symlink name, and goes to activate them?

Hmm. PARTUUID doesn't help you for msdos partitions, right? Not everything's GPT...

kernel 3.7: http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=d3... util-linux 2.24: https://git.kernel.org/cgit/utils/util-linux/util-linux.git/commit/?id=d67cc...

PARTUUID it is, then ;-) -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Arvin Schnell

7 Apr 7 Apr

12:14

On Mon, Mar 30, 2015 at 09:46:14AM +0200, Oliver Neukum wrote:

...

Hi,

I recently experienced a filesystem issue that forced me to use the rescue system and I saw two issues regarding mounting filesystems that seemed bad to them

1) swap space is mounted by UUID That is problematic. If your swap space is corrupted you use mkswap. There practically is no other option. Doing so, however, you change the UUID and your system comes up without swap pretty much without warning.

Use mkswap -U <old-uuid> ... Regards, Arvin -- Arvin Schnell, Senior Software Engineer, Research & Development SUSE Linux GmbH, GF: Felix Imendörffer, Jane Smithard, Jennifer Guild, Dilip Upmanyu, Graham Norton, HRB 21284 (AG Nürnberg) Maxfeldstraße 5 90409 Nürnberg Germany -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Oliver Neukum

13:01

On Tue, 2015-04-07 at 14:14 +0200, Arvin Schnell wrote:

...

On Mon, Mar 30, 2015 at 09:46:14AM +0200, Oliver Neukum wrote:

...
Hi,

I recently experienced a filesystem issue that forced me to use the rescue system and I saw two issues regarding mounting filesystems that seemed bad to them

1) swap space is mounted by UUID That is problematic. If your swap space is corrupted you use mkswap. There practically is no other option. Doing so, however, you change the UUID and your system comes up without swap pretty much without warning.

Use mkswap -U <old-uuid> ...

That works but requires that I mess around with UUIDs. Reagrds Oliver -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

ellanios82

17:26

On 04/07/2015 04:01 PM, Oliver Neukum wrote:

...

On Tue, 2015-04-07 at 14:14 +0200, Arvin Schnell wrote:

...
On Mon, Mar 30, 2015 at 09:46:14AM +0200, Oliver Neukum wrote:

...
Hi,

I recently experienced a filesystem issue that forced me to use the rescue system and I saw two issues regarding mounting filesystems that seemed bad to them

1) swap space is mounted by UUID That is problematic. If your swap space is corrupted you use mkswap. There practically is no other option. Doing so, however, you change the UUID and your system comes up without swap pretty much without warning. Use mkswap -U <old-uuid> ... That works but requires that I mess around with UUIDs.

Reagrds Oliver .................

- perhaps , make an exception , and not to use UUIDs for swap :: for instance :: /dev/sdb3 swap swap defaults 0 0 .................. regards ellan UUIDs -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Jan Engelhardt

17:40

On Tuesday 2015-04-07 19:26, ellanios82 wrote:

...

- perhaps , make an exception , and not to use

UUIDs for swap ::

for instance ::

/dev/sdb3 swap swap defaults 0 0

This is so bad a joke it's not even funny. /dev/sd* are in the class of the most unstable names you can possibly /get. -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Cristian Rodríguez

18:31

On Tue, Apr 7, 2015 at 2:26 PM, ellanios82 wrote:

...

/dev/sdb3 swap swap defaults 0 0

No! device names are not persistent , that's why we are discussing this at all. -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

ellanios82

18:35

On 04/07/2015 08:40 PM, Jan Engelhardt wrote:

...

On Tuesday 2015-04-07 19:26, ellanios82 wrote:

...
- perhaps , make an exception , and not to use

UUIDs for swap ::

for instance ::

/dev/sdb3 swap swap defaults 0 0

This is so bad a joke it's not even funny.

/dev/sd* are in the class of the most unstable names you can possibly /get. .................

- how ethnic : thankyou thankyou thankyou : toujours la politesse ................ -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Yamaban

20:11

New subject: [opensuse-factory] Re: fstab, systemd and UUID

On Tue, 7 Apr 2015 20:31, Cristian Rodríguez wrote:

...

On Tue, Apr 7, 2015 at 2:26 PM, ellanios82 wrote:

...
/dev/sdb3 swap swap defaults 0 0

No! device names are not persistent , that's why we are discussing this at all.

Ok, back to the planning board. Points made (and proven): * "/dev/sd??" names are not persistent, not the solution we seek. * "UUID" as it is, nice for filesystems, not so nice for swap. * "/dev/disk/by-id/(ata|scsi-SATA)_{Dev_Model}_{Dev_Serial}-part?" That is a valid (uniq, save to move port/order) solution, but it is a really LONG string. (Some dislike this approach) How about this mix: We take the "UUID" approach, but change it some. ACSII "SWAP" translates to "53 57 41 50 0a", let this be the first part of any swap Partition UUID. even better if in the next part is the {Dev_Serial} For the full mounty: "SWAP{Dev_Model}{Dev_Serial}{Part.no}{Fillup,random}" convert this -> toHex, use as UUID That way "swapon" can identify a valid swap partition, even if the UUID changes. All it would take is a change in mkswap and swapon code, no other stuff has to be touched. Please discuss. - Yamaban.

Jan Engelhardt

20:54

New subject: [opensuse-factory] Re: fstab, systemd and UUID

On Tuesday 2015-04-07 22:11, Yamaban wrote:

...

On Tue, 7 Apr 2015 20:31, Cristian Rodríguez wrote:

...
On Tue, Apr 7, 2015 at 2:26 PM, ellanios82 wrote:

...
/dev/sdb3 swap swap defaults 0 0

No! device names are not persistent , that's why we are discussing this at all.

How about this mix:

No. We have PARTUUID, and that's working since 3.7 for msdos ptabs if you followed the thread, there is no need to invent a half-dollar custom insular solution.

...

We take the "UUID" approach, but change it some. ACSII "SWAP" translates to "53 57 41 50 0a", let this be the first part of any swap Partition UUID. even better if in the next part is the {Dev_Serial}

For the full mounty: "SWAP{Dev_Model}{Dev_Serial}{Part.no}{Fillup,random}"

This is nonsensical. Not all block devices have a model or serial number.

...

That way "swapon" can identify a valid swap partition,

Whether a partition is valid swap one can figure out by looking at the magic bytes. We do not need to encode SWAP$model$serial$blahblah for that. -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Andrei Borzenkov

8 Apr 8 Apr

03:39

New subject: [opensuse-factory] Re: fstab, systemd and UUID

В Tue, 7 Apr 2015 22:54:15 +0200 (CEST) Jan Engelhardt пишет:

...

On Tuesday 2015-04-07 22:11, Yamaban wrote:

...
On Tue, 7 Apr 2015 20:31, Cristian Rodríguez wrote:

...
On Tue, Apr 7, 2015 at 2:26 PM, ellanios82 wrote:

...
/dev/sdb3 swap swap defaults 0 0

No! device names are not persistent , that's why we are discussing this at all.

How about this mix:

No.

We have PARTUUID, and that's working since 3.7 for msdos

There is no guarantee MS-DOS signature exists so you need some fallback anyway.

...

ptabs if you followed the thread, there is no need to invent a half-dollar custom insular solution.

...
We take the "UUID" approach, but change it some. ACSII "SWAP" translates to "53 57 41 50 0a", let this be the first part of any swap Partition UUID. even better if in the next part is the {Dev_Serial}

For the full mounty: "SWAP{Dev_Model}{Dev_Serial}{Part.no}{Fillup,random}"

This is nonsensical. Not all block devices have a model or serial number.

...
That way "swapon" can identify a valid swap partition,

Whether a partition is valid swap one can figure out by looking at the magic bytes. We do not need to encode SWAP$model$serial$blahblah for that.

-- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Felix Miata

04:45

# blkid /dev/sdb7 /dev/sdb7: LABEL="1r2swapper" UUID="9dbb8c3a-bb2c-49a7-8cdf-85b4dcfd40d1" TYPE="swap" # grep swap /etc/fstab LABEL=1r2swapper swap swap defaults 0 0 When an ordinary human needs to deal with it, labels are far easier than UUIDs. Labels can be short enough to remember at least for a few seconds. -- "The wise are known for their understanding, and pleasant words are persuasive." Proverbs 16:21 (New Living Translation) Team OS/2 ** Reg. Linux User #211409 ** a11y rocks! Felix Miata *** http://fm.no-ip.com/ -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Arvin Schnell

06:55

New subject: [opensuse-factory] Re: fstab, systemd and UUID

On Thu, Apr 02, 2015 at 12:30:16PM +0200, Jan Engelhardt wrote:

...

On Monday 2015-03-30 12:09, Marcos Mello wrote:

...
...
...
...
But the distribution policy is to use UUID We should be opening a FATE request to revert that for swap partitions.

We should be using PARTUUID indeed. But then again, is there not some magic which just automatically goes through all swap devices, irrespective of any device or symlink name, and goes to activate them?

Hmm. PARTUUID doesn't help you for msdos partitions, right? Not everything's GPT...

kernel 3.7: http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=d3... util-linux 2.24: https://git.kernel.org/cgit/utils/util-linux/util-linux.git/commit/?id=d67cc...

PARTUUID it is, then ;-)

That will make the system prone to other modifications: When swap is on partition 6 and you delete partition 5 the number of the swap partition changes to 5 and you again have to update fstab. Regards, Arvin -- Arvin Schnell, Senior Software Engineer, Research & Development SUSE Linux GmbH, GF: Felix Imendörffer, Jane Smithard, Jennifer Guild, Dilip Upmanyu, Graham Norton, HRB 21284 (AG Nürnberg) Maxfeldstraße 5 90409 Nürnberg Germany -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Stefan Seyfried

08:22

Am 07.04.2015 um 19:40 schrieb Jan Engelhardt:

...

On Tuesday 2015-04-07 19:26, ellanios82 wrote:

...
- perhaps , make an exception , and not to use

UUIDs for swap ::

for instance ::

/dev/sdb3 swap swap defaults 0 0

This is so bad a joke it's not even funny.

/dev/sd* are in the class of the most unstable names you can possibly /get.

No, actually not. In "normal" setups (no multipath, fancy storage etc), all disks are /dev/sd* The only exception is KVM Virtualization if people still use the (old, featureless, not useful) virtio-blk instead of the (better, newer, more features) virtio-scsi. The solution there: use virtio-scsi, and disks are again /dev/sd*. And in normal setups, people do usually not reconfigure their built-in disks every day to cause sda->sdc changes, and if they do, they can still choose boot-by-whatever to make this painless. Actually everything but /dev/sdX has been a major PITA, because it has been eiter unstable or not usabe for humans. /dev/disk/by-uuid: seriously, this is for machines, not for humans. /dev/disk/by-id: *not* stable. scsi-1ATA vs scsi-SATA vs ata- anyone? /dev/disk/by-path: *not* stable, at least with virtualization /dev/disk/by-label: *not* unique, better not plug in your old disk in an USB case, I use it anyway but know the pitfalls /dev/sd*: just works most of the time, in standard setups. -- Stefan Seyfried "For a successful technology, reality must take precedence over public relations, for nature cannot be fooled." -- Richard Feynman -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Stefan Seyfried

08:23

Am 07.04.2015 um 20:31 schrieb Cristian Rodríguez:

...

On Tue, Apr 7, 2015 at 2:26 PM, ellanios82 wrote:

...
/dev/sdb3 swap swap defaults 0 0

No! device names are not persistent , that's why we are discussing this at all.

in practice, they are. Much more than all those persistent links which are changing all the time. -- Stefan Seyfried "For a successful technology, reality must take precedence over public relations, for nature cannot be fooled." -- Richard Feynman -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Arvin Schnell

08:42

On Wed, Apr 08, 2015 at 10:23:28AM +0200, Stefan Seyfried wrote:

...

Am 07.04.2015 um 20:31 schrieb Cristian Rodríguez:

...
On Tue, Apr 7, 2015 at 2:26 PM, ellanios82 wrote:

...
/dev/sdb3 swap swap defaults 0 0

No! device names are not persistent , that's why we are discussing this at all.

in practice, they are.

USB disks are something out of practice?

...

Much more than all those persistent links which are changing all the time.

The by-id links have proven to be broken/not persistent. That's why we changed to UUID. Actually, I cannot recall a bug report where UUID failed without the admin doing something stupid. Before switching to UUID we did have bug reports again and again. Regards, Arvin

...

-- Stefan Seyfried

"For a successful technology, reality must take precedence over public relations, for nature cannot be fooled." -- Richard Feynman -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

-- Arvin Schnell, Senior Software Engineer, Research & Development SUSE Linux GmbH, GF: Felix Imendörffer, Jane Smithard, Jennifer Guild, Dilip Upmanyu, Graham Norton, HRB 21284 (AG Nürnberg) Maxfeldstraße 5 90409 Nürnberg Germany -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Oliver Neukum

08:53

On Wed, 2015-04-08 at 10:42 +0200, Arvin Schnell wrote:

...

On Wed, Apr 08, 2015 at 10:23:28AM +0200, Stefan Seyfried wrote:

...
Am 07.04.2015 um 20:31 schrieb Cristian Rodríguez:

...
On Tue, Apr 7, 2015 at 2:26 PM, ellanios82 wrote:

...
/dev/sdb3 swap swap defaults 0 0

No! device names are not persistent , that's why we are discussing this at all.

in practice, they are.

USB disks are something out of practice?

In fstab?

...

...
Much more than all those persistent links which are changing all the time.

The by-id links have proven to be broken/not persistent. That's why we changed to UUID.

Actually, I cannot recall a bug report where UUID failed without the admin doing something stupid. Before switching to UUID we did have bug reports again and again.

They fail the other way round. They are certainly causing me problems. If you like I can make bugzilla reports. Regards Oliver -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Stefan Seyfried

08:56

Am 08.04.2015 um 10:42 schrieb Arvin Schnell:

...

On Wed, Apr 08, 2015 at 10:23:28AM +0200, Stefan Seyfried wrote:

...
Am 07.04.2015 um 20:31 schrieb Cristian Rodríguez:

...
On Tue, Apr 7, 2015 at 2:26 PM, ellanios82 wrote:

...
/dev/sdb3 swap swap defaults 0 0

No! device names are not persistent , that's why we are discussing this at all.

in practice, they are.

USB disks are something out of practice?

in standard setups, those are not mounted via fstab, but via udisks / desktop-environment. The USB disks in your server might be different, but then you are still free to use UUID. And put swap onto them. -- Stefan Seyfried "For a successful technology, reality must take precedence over public relations, for nature cannot be fooled." -- Richard Feynman -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Arvin Schnell

09:43

On Wed, Apr 08, 2015 at 10:56:12AM +0200, Stefan Seyfried wrote:

...

Am 08.04.2015 um 10:42 schrieb Arvin Schnell:

...
On Wed, Apr 08, 2015 at 10:23:28AM +0200, Stefan Seyfried wrote:

...
Am 07.04.2015 um 20:31 schrieb Cristian Rodríguez:

...
On Tue, Apr 7, 2015 at 2:26 PM, ellanios82 wrote:

...
/dev/sdb3 swap swap defaults 0 0

No! device names are not persistent , that's why we are discussing this at all.

in practice, they are.

USB disks are something out of practice?

in standard setups, those are not mounted via fstab, but via udisks / desktop-environment.

...

The USB disks in your server might be different, but then you are still free to use UUID. And put swap onto them.

LOL: You are also free to use whatever you want. If you want to discuss the default options of openSUSE you have to take many use-cases into account - not just yours. Regards, Arvin -- Arvin Schnell, Senior Software Engineer, Research & Development SUSE Linux GmbH, GF: Felix Imendörffer, Jane Smithard, Jennifer Guild, Dilip Upmanyu, Graham Norton, HRB 21284 (AG Nürnberg) Maxfeldstraße 5 90409 Nürnberg Germany -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Stefan Seyfried

09:48

Am 08.04.2015 um 11:43 schrieb Arvin Schnell:

...

On Wed, Apr 08, 2015 at 10:56:12AM +0200, Stefan Seyfried wrote:

...
Am 08.04.2015 um 10:42 schrieb Arvin Schnell:

...
On Wed, Apr 08, 2015 at 10:23:28AM +0200, Stefan Seyfried wrote:

...
Am 07.04.2015 um 20:31 schrieb Cristian Rodríguez:

...
On Tue, Apr 7, 2015 at 2:26 PM, ellanios82 wrote:

...
/dev/sdb3 swap swap defaults 0 0

No! device names are not persistent , that's why we are discussing this at all.

in practice, they are.

USB disks are something out of practice?

in standard setups, those are not mounted via fstab, but via udisks / desktop-environment.

...
The USB disks in your server might be different, but then you are still free to use UUID. And put swap onto them.

LOL: You are also free to use whatever you want.

No, you're actually not. At least not if you're installing with yast. It does work, if you install by booting the rescue-cd, partition your system manually, then use the obs build script to populate the target file sytem, configure everything by hand and reboot afterwards. But it's a bit tedious to do so and if I'd file a droprequest for yast2-installation, I'm quite sure many people would be unhappy. "How to reproduce"? start Installation Partitioner -> expert options -> "mount by device name" (or by whatever, by-label also does not work). Finish installation. Installed system boots by UUID (until a few $TIME_UNITS ago, it did boot by-id or by-path, I'm not sure I remember this correctly).

...

If you want to discuss the default options of openSUSE you have to take many use-cases into account - not just yours.

I'm just not installing anymore, just upgrading. -- Stefan Seyfried "For a successful technology, reality must take precedence over public relations, for nature cannot be fooled." -- Richard Feynman -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Olaf Hering

09:54

On Wed, Apr 08, Stefan Seyfried wrote:

...

Installed system boots by UUID (until a few $TIME_UNITS ago, it did boot by-id or by-path, I'm not sure I remember this correctly).

This is the fault of grub2 and dracut, both ignore the admin choice. Olaf -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Jan Engelhardt

10:15

On Wednesday 2015-04-08 10:22, Stefan Seyfried wrote:

...

...
/dev/sd* are in the class of the most unstable names you can possibly /get.

No, actually not.

In "normal" setups (no multipath, fancy storage etc), all disks are /dev/sd*

I have access to a normal setup. Typical desktop tower box (19"x4RUx~55cm usual size), with extra 3.5" cages ex factory. Contains 16 disks, connected via the cheap €19 off-the-shelf SATA PCIe cards, no "fancy" orderly backplanes, everything is more or less loose the way you'd expect it from a home user. sata_sil and 2x sata_mvw IIRC, and then of course the onboard ahci. Depending on the moon phase, the assignment of sdg-sdp regularly shuffle at boot. -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

ellanios82

13:03

On 04/08/2015 11:22 AM, Stefan Seyfried wrote:
> Am 07.04.2015 um 19:40 schrieb Jan Engelhardt:
>> On Tuesday 2015-04-07 19:26, ellanios82 wrote:
>>> - perhaps , make an exception , and not to use
>>>
>>> UUIDs  for swap ::
>>>
>>> for instance ::
>>>
>>>            /dev/sdb3  swap  swap  defaults   0 0
>>
>> This is so bad a joke it's not even funny.
>>
>> /dev/sd* are in the class of the most unstable names you can possibly
>> /get.
> No, actually not.
>
> In "normal" setups (no multipath, fancy storage etc), all disks are /dev/sd*
>
> The only exception is KVM Virtualization if people still use the (old,
> featureless, not useful) virtio-blk instead of the (better, newer, more
> features) virtio-scsi. The solution there: use virtio-scsi, and disks
> are again /dev/sd*.
>
> And in normal setups, people do usually not reconfigure their built-in
> disks every day to cause sda->sdc changes, and if they do, they can
> still choose boot-by-whatever to make this painless.
>
> Actually everything but /dev/sdX has been a major PITA, because it has
> been eiter unstable or not usabe for humans.
>
> /dev/disk/by-uuid: seriously, this is for machines, not for humans.
> /dev/disk/by-id: *not* stable. scsi-1ATA vs scsi-SATA vs ata- anyone?
> /dev/disk/by-path: *not* stable, at least with virtualization
> /dev/disk/by-label: *not* unique, better not plug in your old disk in an
> USB case, I use it anyway but know the pitfalls
> /dev/sd*: just works most of the time, in standard setups.
.................
  - thankyou, thankyou , thankyou
......................
  regards

-- 
To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org
To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Stefan Seyfried

21:24

Am 08.04.2015 um 12:15 schrieb Jan Engelhardt:

...

On Wednesday 2015-04-08 10:22, Stefan Seyfried wrote:

...
...
/dev/sd* are in the class of the most unstable names you can possibly /get.

No, actually not.

In "normal" setups (no multipath, fancy storage etc), all disks are /dev/sd*

I have access to a normal setup.

Typical desktop tower box (19"x4RUx~55cm usual size), with extra 3.5" cages ex factory. Contains 16 disks,

"normal".

...

connected via the cheap €19 off-the-shelf SATA PCIe cards,

"normal".

...

no "fancy" orderly backplanes, everything is more or less loose the way you'd expect it from a home user. sata_sil and 2x sata_mvw IIRC, and then of course the onboard ahci.

Depending on the moon phase, the assignment of sdg-sdp regularly shuffle at boot.

I'd guess this happens after kernel updates or such, not when rebooting the same kernel without rebuilding initrd. But anyway, these are the cases where /dev/disk/by-$WHATEVER is made for. I still would use by-label, but you can use anything and it will work. a "normal" setup as I consider it today allows you to only put in one disk. Everything more is external (USB, ExpressCard or such) and probably not put into fstab anyway. A "normal" setup nowadays (IMHO) for "normal" users is a notebook computer, maybe a nettop or stuff like the intel NUC boxes. People building home storage arrays with 16 disks, without using LVM or raid setups, are usually able to configure by-$WHATEVER. I too still have a single "home server" box with 3x rotating rust plus 1 ssd (used for bcache) plus DVB Cards plus serial multiport cards. And I actually can make the disks change order by changing BIOS settings. But I wouldn't dare using this example for advocating default for standard installations. And it mounts everything "by-label", except stuff that's on LVM: seife@server:~> grep -v ^# /etc/fstab LABEL=root / ext4 acl,user_xattr 1 1 LABEL=boot /boot ext3 acl,user_xattr 1 1 LABEL=log-journal /var/log/journal xfs defaults 1 1 LABEL=home.bcache /home ext4 acl,user_xattr 1 2 LABEL=space1.bcach /space1 xfs logbufs=8,logbsize=262144 1 2 /dev/local/system-swap swap swap defaults 0 0 /dev/4TB.1/4TB.1 /space2 xfs logbufs=8,logbsize=256k,inode64 1 2 LABEL=2TB_space3 /space3 xfs logbufs=8,logbsize=262144 1 2 LABEL=srv /srv xfs logbufs=8,logbsize=262144 1 2 /space2/capture/vdr /export/vdr fuse.vdrfs allow_other 0 0 /space1/dist/openSUSE-13.2-DVD-Build0019-x86_64.iso /space1/dist/openSUSE-13.2-x86_64 iso9660 loop,ro 0 0 [...snip /proc,/sys, whatever is not a device...] seife@server:~> cat /proc/cmdline root=/dev/disk/by-label/root sysrq=yes printk.time=1 acpi_enforce_resources=lax -- -- Stefan Seyfried "For a successful technology, reality must take precedence over public relations, for nature cannot be fooled." -- Richard Feynman "Your mail is 7 pages of printout. Do you seriously expect people that do openSUSE in their free time to read that? Little less Castro, little more JFK..." -- coolo -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Dominique Leuenberger / DimStar

21:28

On Wed, 2015-04-08 at 23:24 +0200, Stefan Seyfried wrote:

...

a "normal" setup as I consider it today allows you to only put in one disk. Everything more is external (USB, ExpressCard or such) and probably not put into fstab anyway.

Serious? one disk only? My notebook already has two disks (one SSD and one rotating)... so this system already is not considered 'normal'. I guess we need to change the scope of normal a bit more than 'Aldi- Computers' Dominique -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Stefan Seyfried

21:37

Am 08.04.2015 um 11:54 schrieb Olaf Hering:

...

On Wed, Apr 08, Stefan Seyfried wrote:

...
Installed system boots by UUID (until a few $TIME_UNITS ago, it did boot by-id or by-path, I'm not sure I remember this correctly).

This is the fault of grub2 and dracut, both ignore the admin choice.

Last time I looked, yast did, too (installed fstab was wrong). The magic is: you need to change your choice, then delete all partitioning and recreate manually. Your choice will not affect the proposal that was created before you were even able to choose. Totally logical from a programmers point of view. Totally useless from a users point of view. Reminds me of a famous quote attributed to A'rpi (of MPlayer fame): "Users? My software runs just fine without users!" ;-) -- Stefan Seyfried "For a successful technology, reality must take precedence over public relations, for nature cannot be fooled." -- Richard Feynman -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Stefan Seyfried

21:43

Am 08.04.2015 um 23:28 schrieb Dominique Leuenberger / DimStar:

...

On Wed, 2015-04-08 at 23:24 +0200, Stefan Seyfried wrote:

...
a "normal" setup as I consider it today allows you to only put in one disk. Everything more is external (USB, ExpressCard or such) and probably not put into fstab anyway.

Serious? one disk only? My notebook already has two disks (one SSD and one rotating)... so this system already is not considered 'normal'.

Can you make them change device names? Means: multiple controllers, driven by multiple drivers or such? If no: standard setup. All notebooks I know of which have multiple disks still only have one disk controller type and thus a deterministic order of device discovery.

...

I guess we need to change the scope of normal a bit more than 'Aldi- Computers'

Actually the Aldi stuff might be the ones most likely violating my assumptions, but fortunately the audience of those is buying tablets nowadays, so theses are going to vanish quite soon :-) -- Stefan Seyfried "For a successful technology, reality must take precedence over public relations, for nature cannot be fooled." -- Richard Feynman -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Jan Engelhardt

21:47

On Wednesday 2015-04-08 23:24, Stefan Seyfried wrote:

...

...
...
...
/dev/sd* are in the class of the most unstable names you can possibly get.

No, actually not. In "normal" setups (no multipath, fancy storage etc), all disks are /dev/sd*

I have access to a normal setup. Typical desktop tower box (19"x4RUx~55cm usual size), with extra 3.5" cages ex factory. Contains 16 disks,

"normal".

Yes, it _is_ normal.

...

a "normal" setup as I consider it today allows you to only put in one disk.

Quite a narrow definition of "normal" you have there. I say normal is what's within 2 sigma. Boring one-controller setups are within that 2 sigma, but are not a majority within it, for server role setups are still quite strong, and I hear we are still waiting for The Year Of The Linux Desktop. -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Stefan Seyfried

21:57

Am 08.04.2015 um 23:47 schrieb Jan Engelhardt:

...

On Wednesday 2015-04-08 23:24, Stefan Seyfried wrote:

...
...
I have access to a normal setup. Typical desktop tower box (19"x4RUx~55cm usual size), with extra 3.5" cages ex factory. Contains 16 disks,

"normal".

Yes, it _is_ normal.

If you say it.

...

...
a "normal" setup as I consider it today allows you to only put in one disk.

Quite a narrow definition of "normal" you have there. I say normal is what's within 2 sigma.

The question is: what setup should be catered for with the default parameters.

...

Boring one-controller setups are within that 2 sigma, but are not a majority within it, for server role setups are still quite strong,

Oh yes. And server setups with more than one, maybe two controllers, as everyone is using local disks on servers (for anything else than booting the OS). Certainly. I have yet to find those in the "few" servers at work.

...

and I hear we are still waiting for The Year Of The Linux Desktop.

I rest my case. * I don't do fresh installs anymore anyway. * No matter what you choose in Yast, it is not respected anyway. So go ahead, fix all those 17-disks-attached-to-15-controllers cases so that they work fine with the default settings. And make sure, that the boring one-controller setups (as well as the equally boring two-controllers-of-the-same-type setups) are breaking subtly all the time as they are now. Sounds like a plan. -- Stefan Seyfried "For a successful technology, reality must take precedence over public relations, for nature cannot be fooled." -- Richard Feynman -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Oliver Neukum

9 Apr 9 Apr

06:38

On Wed, 2015-04-08 at 23:43 +0200, Stefan Seyfried wrote:

...

If no: standard setup. All notebooks I know of which have multiple disks still only have one disk controller type and thus a deterministic order of device discovery.

That we definitely cannot assume. A rotational SATA disk and an SSD over NVME or another interface not ATA are pretty common in larger laptops under development. SATA is just not sensible interface for integrated circuits. Regards Oliver -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Arvin Schnell

06:42

On Wed, Apr 08, 2015 at 11:37:59PM +0200, Stefan Seyfried wrote:

...

Am 08.04.2015 um 11:54 schrieb Olaf Hering:

...
On Wed, Apr 08, Stefan Seyfried wrote:

...
Installed system boots by UUID (until a few $TIME_UNITS ago, it did boot by-id or by-path, I'm not sure I remember this correctly).

This is the fault of grub2 and dracut, both ignore the admin choice.

Last time I looked, yast did, too (installed fstab was wrong).

The magic is: you need to change your choice, then delete all partitioning and recreate manually. Your choice will not affect the proposal that was created before you were even able to choose. Totally logical from a programmers point of view. Totally useless from a users point of view.

And the feature request to allow regenerating the proposal has been postponed for several years (or even dropped by now). So please stop blaming developers here. ciao Arvin -- Arvin Schnell, Senior Software Engineer, Research & Development SUSE Linux GmbH, GF: Felix Imendörffer, Jane Smithard, Jennifer Guild, Dilip Upmanyu, Graham Norton, HRB 21284 (AG Nürnberg) Maxfeldstraße 5 90409 Nürnberg Germany -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Stefan Seyfried

18:30

Am 09.04.2015 um 08:38 schrieb Oliver Neukum:

...

On Wed, 2015-04-08 at 23:43 +0200, Stefan Seyfried wrote:

...
If no: standard setup. All notebooks I know of which have multiple disks still only have one disk controller type and thus a deterministic order of device discovery.

That we definitely cannot assume. A rotational SATA disk and an SSD over NVME or another interface not ATA are pretty common in larger laptops under development.

They are still unlikely to switch device names randomly on every boot. But as I wrote -- I'm not interested in this problem anymore, but will laugh everytime a bug with this setup is reported and tell people that it is by design. -- Stefan Seyfried "For a successful technology, reality must take precedence over public relations, for nature cannot be fooled." -- Richard Feynman -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

skylar

18:40

I honestly don’t know why anybody uses swap anymore anyway.... On 04/09/2015 12:30 PM, Stefan Seyfried wrote:

...

Am 09.04.2015 um 08:38 schrieb Oliver Neukum:

...
On Wed, 2015-04-08 at 23:43 +0200, Stefan Seyfried wrote:

...
If no: standard setup. All notebooks I know of which have multiple disks still only have one disk controller type and thus a deterministic order of device discovery. That we definitely cannot assume. A rotational SATA disk and an SSD over NVME or another interface not ATA are pretty common in larger laptops under development. They are still unlikely to switch device names randomly on every boot.

But as I wrote -- I'm not interested in this problem anymore, but will laugh everytime a bug with this setup is reported and tell people that it is by design.

-- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Rüdiger Meier

18:50

On April 9, 2015 8:40:36 PM CEST, skylar wrote:

...

I honestly don’t know why anybody uses swap anymore anyway....

Since there are many tmpfs mounts per default nowadays, swap is a must again IMO. -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Jan Engelhardt

18:59

On Thursday 2015-04-09 20:50, Rüdiger Meier wrote:

...

On April 9, 2015 8:40:36 PM CEST, skylar wrote:

...
I honestly don’t know why anybody uses swap anymore anyway....

Since there are many tmpfs mounts per default nowadays, swap is a must again IMO.

Yeah, though swap partitions are an unlikely answer (they often go unused). A better space utilization would be something like making the affected tmpfs directories be mounts of a volatile btrfs subvolume instead. -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Oliver Neukum

20:45

On Thu, 2015-04-09 at 12:40 -0600, skylar wrote:

...

I honestly don’t know why anybody uses swap anymore anyway....

S4 Regards Oliver -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Carlos E. R.

11 Apr 11 Apr

21:10

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA256 On 2015-04-08 10:22, Stefan Seyfried wrote:

...

Am 07.04.2015 um 19:40 schrieb Jan Engelhardt:

...

...
/dev/sd* are in the class of the most unstable names you can possibly /get.

No, actually not.

In "normal" setups (no multipath, fancy storage etc), all disks are /dev/sd*

The only exception is KVM Virtualization if people still use the (old, featureless, not useful) virtio-blk instead of the (better, newer, more features) virtio-scsi. The solution there: use virtio-scsi, and disks are again /dev/sd*.

In my desktop, /dev/sda can become sdb or sdc on the next boot, if I happened to hotplug some other disk (maybe usb, maybe esata). Using of /dev/sd* names has been deprecated for a reason. - -- Cheers / Saludos, Carlos E. R. (from 13.1 x86_64 "Bottle" (Minas Tirith)) -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.22 (GNU/Linux) iF4EAREIAAYFAlUpjcQACgkQja8UbcUWM1x9PAD/fzNVLXHVMCXx47FWwliywBoZ dGsunL5RXYptwZ6zsRAA/jmTS9z3n7IYBpvFMdy8nC2axYYsT0vtwYKPV7t4Kuvi =qf2w -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Carlos E. R.

21:21

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA256 On 2015-04-08 23:57, Stefan Seyfried wrote:

...

The question is: what setup should be catered for with the default parameters.

All. - -- Cheers / Saludos, Carlos E. R. (from 13.1 x86_64 "Bottle" (Minas Tirith)) -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.22 (GNU/Linux) iF4EAREIAAYFAlUpkFEACgkQja8UbcUWM1zEdwD/c7cUXysmLj9nWe8NoFMrVI5U q+PZ1AogycU4Xl04T0MA/3G3zRCG/cFENtfRA8UXiidz3wQ4wN5apCSlabSjD+kT =eSUV -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Carlos E. R.

21:28

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA256 On 2015-04-08 10:53, Oliver Neukum wrote:

...

On Wed, 2015-04-08 at 10:42 +0200, Arvin Schnell wrote:

...

...
USB disks are something out of practice?

In fstab?

I do. For backups and archival. Yes, in fstab. - -- Cheers / Saludos, Carlos E. R. (from 13.1 x86_64 "Bottle" (Minas Tirith)) -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.22 (GNU/Linux) iF4EAREIAAYFAlUpkgYACgkQja8UbcUWM1w1sgD/Tfoub2bErpMWJo9Rw39qBDMO u3tMl4UMpMQHKcZlymEBAIhbxpLVpQgPqw6nYt62jbpaKN81403FrwLLoqg8z1wl =9bRA -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Carlos E. R.

21:30

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA256 On 2015-04-08 11:43, Arvin Schnell wrote:

...

On Wed, Apr 08, 2015 at 10:56:12AM +0200, Stefan Seyfried wrote:

...

If you want to discuss the default options of openSUSE you have to take many use-cases into account - not just yours.

Exactly. - -- Cheers / Saludos, Carlos E. R. (from 13.1 x86_64 "Bottle" (Minas Tirith)) -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.22 (GNU/Linux) iF4EAREIAAYFAlUpknUACgkQja8UbcUWM1yQpQD/cAmz/zf2j4ax3zh0mPgu882t d8ytHsxlgBfMx4JLZ7YBAJZssKVarIuwpj8lcBXzbt/QD6L+Vjp4iNqUoKD9KtEf =G3wY -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

3303

Age (days ago)

3315

Last active (days ago)

List overview

Download

52 comments

17 participants

participants (17)

Andrei Borzenkov
Arvin Schnell
Carlos E. R.
Carlos E. R.
Cristian Rodríguez
Dominique Leuenberger / DimStar
ellanios82
Felix Miata
Hannes Reinecke
Jan Engelhardt
Marcos Mello
Olaf Hering
Oliver Neukum
Rüdiger Meier
skylar
Stefan Seyfried
Yamaban