[opensuse-factory] ext3 external journal impossible?
Anybody an idea why (unlike with xfs or reiserfs) I could not manage to make an ext3 filesystem with external journal. After "mkfs.ext3 -b 4096 -O /dev/sda5 /dev/hda7" I get: "Unvollständige Dateisystem-Option gesetzt: /dev/sda5" And after "mke2fs -b 4096 -n -j -J device=/dev/hda7 /dev/sda5" I end up with "Filesystem has unexpected block size beim Versuch, das Journal-Device /dev/hda7 zu öffnen" at the very end of the process. Also using tune2fs to convert: little:/home/fmf # tune2fs -j /dev/sda5 -J device=/dev/hda7 tune2fs 1.39 (29-May-2006) tune2fs: Filesystem has unexpected block size beim Öffnen des Journals auf /dev/hda7 after making an ext2 filesystem fails. What did I overlook here? Appreciating any help FMF --------------------------------------------------------------------- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
On Sunday 26 November 2006 11:06, Frank-Michael Fischer wrote:
Anybody an idea why (unlike with xfs or reiserfs) I could not manage to make an ext3 filesystem with external journal. After "mkfs.ext3 -b 4096 -O /dev/sda5 /dev/hda7" I get: "Unvollständige Dateisystem-Option gesetzt: /dev/sda5"
Did you create the journal device first, as the man page says? If you want to have the file system on /dev/sda7 and the journal on /dev/sda5, you would do mkfs.ext3 -O journal_dev /dev/sda5 Then you can do mkfs.ext3 -j -J device=/dev/sda5 /dev/sda7 --------------------------------------------------------------------- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
Anders Johansson schrieb:
On Sunday 26 November 2006 11:06, Frank-Michael Fischer wrote:
Anybody an idea why (unlike with xfs or reiserfs) I could not manage to make an ext3 filesystem with external journal. After "mkfs.ext3 -b 4096 -O /dev/sda5 /dev/hda7" I get: "Unvollständige Dateisystem-Option gesetzt: /dev/sda5"
Did you create the journal device first, as the man page says?
If you want to have the file system on /dev/sda7 and the journal on /dev/sda5, you would do
mkfs.ext3 -O journal_dev /dev/sda5
Then you can do
mkfs.ext3 -j -J device=/dev/sda5 /dev/sda7 According to my message, this is what I did:
mkfs.ext3 -b 4096 -O /dev/sda5 /dev/hda7 and then mke2fs -b 4096 -n -j -J device=/dev/hda7 /dev/sda5 what should I change? Is there anybody who managed to create ext3 with external journal in 10.2 RC1? FMF --------------------------------------------------------------------- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
On Sunday 26 November 2006 11:24, Frank-Michael Fischer wrote:
Anders Johansson schrieb:
On Sunday 26 November 2006 11:06, Frank-Michael Fischer wrote:
Anybody an idea why (unlike with xfs or reiserfs) I could not manage to make an ext3 filesystem with external journal. After "mkfs.ext3 -b 4096 -O /dev/sda5 /dev/hda7" I get: "Unvollständige Dateisystem-Option gesetzt: /dev/sda5"
Did you create the journal device first, as the man page says?
If you want to have the file system on /dev/sda7 and the journal on /dev/sda5, you would do
mkfs.ext3 -O journal_dev /dev/sda5
Then you can do
mkfs.ext3 -j -J device=/dev/sda5 /dev/sda7
According to my message, this is what I did:
mkfs.ext3 -b 4096 -O /dev/sda5 /dev/hda7
I think you misunderstand the manpage. It should literally be journal_dev, that shouldn't be replaced with a device name. If you want to have the journal on hda7, the command should be mkfs.ext3 -b 4096 -O journal_dev /dev/hda7 --------------------------------------------------------------------- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
Anders Johansson schrieb:
I think you misunderstand the manpage. It should literally be journal_dev, that shouldn't be replaced with a device name.
If you want to have the journal on hda7, the command should be
mkfs.ext3 -b 4096 -O journal_dev /dev/hda7 --------------------------------------------
Right you are, thanks for the help, working now here! FMF --------------------------------------------------------------------- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
Anders Johansson schrieb:
I think you misunderstand the manpage. It should literally be journal_dev, that shouldn't be replaced with a device name.
If you want to have the journal on hda7, the command should be
mkfs.ext3 -b 4096 -O journal_dev /dev/hda7
Interesting: now after external journalling working I found out that using /dev/ram0 as journal device slows writing down a bit compared to local journal or journal on another disk. Strange. FMF --------------------------------------------------------------------- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
On Sunday 26 November 2006 12:33, Frank-Michael Fischer wrote:
Anders Johansson schrieb:
I think you misunderstand the manpage. It should literally be journal_dev, that shouldn't be replaced with a device name.
If you want to have the journal on hda7, the command should be
mkfs.ext3 -b 4096 -O journal_dev /dev/hda7
Interesting: now after external journalling working I found out that using /dev/ram0 as journal device slows writing down a bit compared to local journal or journal on another disk. Strange.
Using a ram disk as journal device is totally and utterly pointless. You might as well disable journalling completely then. The whole idea of keeping a journal is to have a log of metadata changes in case of a crash. If this log is kept in ram, it will be gone, making it completely useless I can't say why it should be slower, but since it shouldn't be used at all, it doesn't seem to matter much --------------------------------------------------------------------- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
On Sun, Nov 26, 2006 at 01:13:45PM +0100, Anders Johansson wrote:
I can't say why it should be slower, but since it shouldn't be used at all, it doesn't seem to matter much
That depends on your point of view. From a user's point of view this is really pointless because it does not make any sense at all to use such a scenario on any productive system. For the developer of the code such observations could be quite valuable because the process of finding an explanation for such strange observations often uncover obscure bugs or design issues. Robert -- Robert Schiele Dipl.-Wirtsch.informatiker mailto:rschiele@gmail.com "Quidquid latine dictum sit, altum sonatur."
On Sunday 26 November 2006 14:01, Robert Schiele wrote:
On Sun, Nov 26, 2006 at 01:13:45PM +0100, Anders Johansson wrote:
I can't say why it should be slower, but since it shouldn't be used at all, it doesn't seem to matter much
That depends on your point of view. From a user's point of view this is really pointless because it does not make any sense at all to use such a scenario on any productive system. For the developer of the code such observations could be quite valuable because the process of finding an explanation for such strange observations often uncover obscure bugs or design issues.
It could be as simple as the RAM disk taking up memory that would otherwise be used for buffering. Reducing the size of the buffers can cause substantial performance loss --------------------------------------------------------------------- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
On Sun, Nov 26, 2006 at 02:04:17PM +0100, Anders Johansson wrote:
On Sunday 26 November 2006 14:01, Robert Schiele wrote:
On Sun, Nov 26, 2006 at 01:13:45PM +0100, Anders Johansson wrote:
I can't say why it should be slower, but since it shouldn't be used at all, it doesn't seem to matter much
That depends on your point of view. From a user's point of view this is really pointless because it does not make any sense at all to use such a scenario on any productive system. For the developer of the code such observations could be quite valuable because the process of finding an explanation for such strange observations often uncover obscure bugs or design issues.
It could be as simple as the RAM disk taking up memory that would otherwise be used for buffering. Reducing the size of the buffers can cause substantial performance loss
Sure this might be an explanation depending on the detailed setup that was done here. But only explanations that you can _prove_ to be the cause of the observation can eventually help uncovering a bug. Just listing explanations that _might_ be a cause could make a Sunday afternoon more entertaining but don't help in any way in the respect of identifying problems. Actually my point was not to force you providing an explanation but just to state that you cannot claim an observation pointless just because it results from a setup that does not make sense for production usage. Robert -- Robert Schiele Dipl.-Wirtsch.informatiker mailto:rschiele@gmail.com "Quidquid latine dictum sit, altum sonatur."
Anders Johansson schrieb:
On Sunday 26 November 2006 14:01, Robert Schiele wrote:
On Sun, Nov 26, 2006 at 01:13:45PM +0100, Anders Johansson wrote:
I can't say why it should be slower, but since it shouldn't be used at all, it doesn't seem to matter much
That depends on your point of view. From a user's point of view this is really pointless because it does not make any sense at all to use such a scenario on any productive system. For the developer of the code such observations could be quite valuable because the process of finding an explanation for such strange observations often uncover obscure bugs or design issues.
It could be as simple as the RAM disk taking up memory that would otherwise be used for buffering. Reducing the size of the buffers can cause substantial performance loss Not really in my case, only about 50% of memory in use including the 32MB ram disk. FMF
To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
Robert Schiele schrieb:
On Sun, Nov 26, 2006 at 01:13:45PM +0100, Anders Johansson wrote:
I can't say why it should be slower, but since it shouldn't be used at all, it doesn't seem to matter much
That depends on your point of view. From a user's point of view this is really pointless because it does not make any sense at all to use such a scenario on any productive system. For the developer of the code such observations could be quite valuable because the process of finding an explanation for such strange observations often uncover obscure bugs or design issues.
Robert
Right, I wanted to find out the impact of journal device speed on overall performance and had to learn: there is no useful correlation.There is not even a reasonable performance difference between internal and external journal on ext3. At least none that bonnie would show. FMF --------------------------------------------------------------------- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
participants (3)
-
Anders Johansson -
Frank-Michael Fischer -
Robert Schiele