Bug 133428 - EXT3: Enable data=writeback on / filesystem in YAST2 installer
Summary: EXT3: Enable data=writeback on / filesystem in YAST2 installer
Status: RESOLVED FIXED
Alias: None
Product: SUSE LINUX 10.0
Classification: openSUSE
Component: YaST2 (show other bugs)
Version: Final
Hardware: Other SuSE Linux 10.0
: P5 - None : Enhancement
Target Milestone: ---
Assignee: Thomas Fehr
QA Contact: Klaus Kämpf
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2005-11-11 10:42 UTC by Jens Siebert
Modified: 2006-08-02 11:19 UTC (History)
1 user (show)

See Also:
Found By: Other
Services Priority:
Business Priority:
Blocker: ---
Marketing QA Status: ---
IT Deployment: ---


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description Jens Siebert 2005-11-11 10:42:03 UTC
In ordered jounalling mode of ext3 data is always written to disk prior to the metadata (journal data). This ensures data consistency in most cases. However the performance of this option is not that well.
In writeback journalling mode the metadata might get comitted to the journal before the data is actually written to the disk. This might lead to data corruption in some cases (e.g. power-off before the data could be written to disk but the journal data already has been commited). But on the other hand performance of this option is considerably better.
The YAST2 installer allows to set the writeback journalling mode but not in the / filesystem. If one specifies the data=writeback option in /etc/fstab the / filesystem is mounted in ordered journalling mode although. To get around this one can use the tune2fs tools with the option '-o journal_data_writeback'. This sets the default mount option for the filesystem with which it gets mounted in writeback journalling mode. The same applies to the 'journal' journalling mode.

So my suggestion for the YAST2 installer is to call' tune2fs -o journal_data_writeback' for the respective device if a user selects the writeback journalling mode on ext3 for the / filesystem. I've tested this on three different machines without any problems.
Comment 1 Michael Gross 2005-11-11 13:21:21 UTC
This might also have a special reason. I think it is very important that the journal is always consistent... that's the whole clue about such a filesystem. When changing this, there should be at least some warning about this. Let's the maintainer decide. A also took Chris into CC here.
Comment 2 Thomas Fehr 2005-11-11 14:08:13 UTC
This is something for PM or dist team to decide.
Comment 3 Andreas Jaeger 2006-01-04 10:01:08 UTC
Thomas, let's allow this for / as well but leave the default as is.
Comment 4 Thomas Fehr 2006-01-09 11:48:33 UTC
Will consider that for SL 10.2
Comment 5 Thomas Fehr 2006-07-13 11:36:55 UTC
Reopen for work in SL 10.2
Comment 6 Thomas Fehr 2006-08-02 11:19:43 UTC
Implemented in current SVN head.