Hi all,
I have what might be a foolish question about patching packages. I am not sure exactly how to phrase the question, so please follow up if it seems as though I'm not being clear.
I was looking at this bug which my machines are currently experiencing:
https://bugzilla.redhat.com/show_bug.cgi?id=883905
The proposed patch is literally one new line in the XFS codebase. So since the patch is so straightforward, I had a crazy idea that I would build my own kernel with this patch, and test it out to see if it worked. (It's been many years since I built my own kernel, so that would be an adventure in and of itself.)
My worry would be, I would want to make sure that I propagated this patch every time I updated the kernel package. Is this something others do regularly? If so, is there a standard way of managing the process of applying one's own patches to a series of source packages, and being able to re-patch and rebuild updated packages?
I'm guessing that I just need to build xfs.ko. Are there any gotchas beyond the wiki entry on building your own kernel modules? That page seems to target CentOS 5, not 6, but I imagine the process is quite similar.
--keith
On Wed, Mar 6, 2013 at 3:22 PM, Keith Keller kkeller@wombat.san-francisco.ca.us wrote:
Hi all,
I have what might be a foolish question about patching packages. I am not sure exactly how to phrase the question, so please follow up if it seems as though I'm not being clear.
I was looking at this bug which my machines are currently experiencing:
https://bugzilla.redhat.com/show_bug.cgi?id=883905
The proposed patch is literally one new line in the XFS codebase. So since the patch is so straightforward, I had a crazy idea that I would build my own kernel with this patch, and test it out to see if it worked. (It's been many years since I built my own kernel, so that would be an adventure in and of itself.)
You may want to check this out:
http://bugs.centos.org/view.php?id=6087
My understanding is that "There is no side effect other than the load. There are not performance issues with the ailds behaving like this." Is this not the case ?
My worry would be, I would want to make sure that I propagated this patch every time I updated the kernel package. Is this something others do regularly? If so, is there a standard way of managing the process of applying one's own patches to a series of source packages, and being able to re-patch and rebuild updated packages?
I'm guessing that I just need to build xfs.ko. Are there any gotchas beyond the wiki entry on building your own kernel modules? That page seems to target CentOS 5, not 6, but I imagine the process is quite similar.
--keith
The wiki article:
http://wiki.centos.org/HowTos/BuildingKernelModules
may not be quite up-to-date in that it does not reflect the kernel version for CentOS 6 (2.6.32). But the principle is there. For building your own modules, you can also download one of the kmod packages from ELRepo and study how it's done.
Akemi
On 2013-03-07, Akemi Yagi amyagi@gmail.com wrote:
You may want to check this out:
http://bugs.centos.org/view.php?id=6087
My understanding is that "There is no side effect other than the load. There are not performance issues with the ailds behaving like this." Is this not the case ?
As far as I can tell, it is. I actually prompted Dave's quoted comment on the XFS list:
http://oss.sgi.com/archives/xfs/2012-11/msg00594.html
So this would be a low priority task for me (as well as a learning exercise). If the patch were two lines I probably wouldn't bother. ;-) It is 99.5% cosmetic, but I have noticed that the ''baseline'' load, when there is no I/O, varies between 3 and 4, which makes it very slightly more difficult to interpret the load. That is my main motivation for bothering--if the baseline were more stable I probably wouldn't bother. (With fewer XFS filesystems mounted the issue is even less obvious.)
The wiki article:
http://wiki.centos.org/HowTos/BuildingKernelModules
may not be quite up-to-date in that it does not reflect the kernel version for CentOS 6 (2.6.32). But the principle is there. For building your own modules, you can also download one of the kmod packages from ELRepo and study how it's done.
Perfect, thank you! If people are interested, and I do make the attempt, I will post my results.
--keith
On Wed, Mar 6, 2013 at 4:25 PM, Keith Keller kkeller@wombat.san-francisco.ca.us wrote:
On 2013-03-07, Akemi Yagi amyagi@gmail.com wrote:
You may want to check this out:
http://bugs.centos.org/view.php?id=6087
My understanding is that "There is no side effect other than the load. There are not performance issues with the ailds behaving like this." Is this not the case ?
As far as I can tell, it is. I actually prompted Dave's quoted comment on the XFS list:
http://oss.sgi.com/archives/xfs/2012-11/msg00594.html
So this would be a low priority task for me (as well as a learning exercise). If the patch were two lines I probably wouldn't bother. ;-) It is 99.5% cosmetic, but I have noticed that the ''baseline'' load, when there is no I/O, varies between 3 and 4, which makes it very slightly more difficult to interpret the load. That is my main motivation for bothering--if the baseline were more stable I probably wouldn't bother. (With fewer XFS filesystems mounted the issue is even less obvious.)
I thought about applying the patch to the centosplus kernel but decided not to bother because it looked like a "non-issue". But it you think it's worth the fix, that can be done. It will be even better if you supply the actual patch for the CentOS kernel.
Akemi
On 2013-03-07, Akemi Yagi amyagi@gmail.com wrote:
I thought about applying the patch to the centosplus kernel but decided not to bother because it looked like a "non-issue". But it you think it's worth the fix, that can be done. It will be even better if you supply the actual patch for the CentOS kernel.
Honestly I do not think it's worth the fix for the centosplus kernel. But perhaps others disagree with me; if so, is the proposed patch by RedHat usable?
https://bugzilla.redhat.com/attachment.cgi?id=674895
If it's not, let me know, and I will try to generate the correct one.
--keith
On Wed, Mar 6, 2013 at 8:16 PM, Keith Keller kkeller@wombat.san-francisco.ca.us wrote:
On 2013-03-07, Akemi Yagi amyagi@gmail.com wrote:
I thought about applying the patch to the centosplus kernel but decided not to bother because it looked like a "non-issue". But it you think it's worth the fix, that can be done. It will be even better if you supply the actual patch for the CentOS kernel.
Honestly I do not think it's worth the fix for the centosplus kernel. But perhaps others disagree with me; if so, is the proposed patch by RedHat usable?
https://bugzilla.redhat.com/attachment.cgi?id=674895
If it's not, let me know, and I will try to generate the correct one.
--keith
I looked at the patch again -- it was apparently not for the 6.3 kernel and I was not 100% sure about applying it at that time. But now with the 6.4 kernel, it cleanly applies.
While not serious, some users are finding this bug to be problematic when monitoring the stats. I will include the fix in the next plus kernel update.
Akemi
On 2013-03-08, Akemi Yagi amyagi@gmail.com wrote:
I looked at the patch again -- it was apparently not for the 6.3 kernel and I was not 100% sure about applying it at that time. But now with the 6.4 kernel, it cleanly applies.
While not serious, some users are finding this bug to be problematic when monitoring the stats. I will include the fix in the next plus kernel update.
That's great, thanks! I'll keep my eye on the repo.
In previous versions, the centosplus additions were documented in a readme file at the top level of the centosplus directory. Are there similar docs for CentOS 6? I checked the wiki and the ftp sites but didn't find anything yet.
--keith
On Fri, Mar 8, 2013 at 2:11 PM, Keith Keller kkeller@wombat.san-francisco.ca.us wrote:
On 2013-03-08, Akemi Yagi amyagi@gmail.com wrote:
I looked at the patch again -- it was apparently not for the 6.3 kernel and I was not 100% sure about applying it at that time. But now with the 6.4 kernel, it cleanly applies.
While not serious, some users are finding this bug to be problematic when monitoring the stats. I will include the fix in the next plus kernel update.
That's great, thanks! I'll keep my eye on the repo.
In previous versions, the centosplus additions were documented in a readme file at the top level of the centosplus directory. Are there similar docs for CentOS 6? I checked the wiki and the ftp sites but didn't find anything yet.
The only place the plus kernels are "documented" with up-to-date info is the CentOS bug tracker:
CentOS-6: http://bugs.centos.org/view.php?id=4586 CentOS-5: http://bugs.centos.org/view.php?id=4100
Akemi
On 2013-03-08, Akemi Yagi amyagi@gmail.com wrote:
The only place the plus kernels are "documented" with up-to-date info is the CentOS bug tracker:
CentOS-6: http://bugs.centos.org/view.php?id=4586 CentOS-5: http://bugs.centos.org/view.php?id=4100
That's perfect. It seems completely reasonable to suggest to people who want to fix this issue to use the centosplus kernel when it comes out instead; I don't see anything that would block most systems from deploying it. Of course I will still try it on a non-critical system first. :)
--keith
On 2013-03-08, Akemi Yagi amyagi@gmail.com wrote:
The only place the plus kernels are "documented" with up-to-date info is the CentOS bug tracker:
CentOS-6: http://bugs.centos.org/view.php?id=4586
At least for CentOS 6 it looks like the new centosplus kernel is out. Thanks Akemi! I have it running on a test box, and it looks like it properly addresses the XFS aild issue. If you want to address this issue and don't feel like managing dkms, it might be worth your while to test out the centosplus kernel on your own machines.
--keith
On Wed, Mar 13, 2013 at 12:24 PM, Keith Keller kkeller@wombat.san-francisco.ca.us wrote:
On 2013-03-08, Akemi Yagi amyagi@gmail.com wrote:
The only place the plus kernels are "documented" with up-to-date info is the CentOS bug tracker:
CentOS-6: http://bugs.centos.org/view.php?id=4586
At least for CentOS 6 it looks like the new centosplus kernel is out. Thanks Akemi! I have it running on a test box, and it looks like it properly addresses the XFS aild issue. If you want to address this issue and don't feel like managing dkms, it might be worth your while to test out the centosplus kernel on your own machines.
--keith
Thanks for reporting back with the confirmation that the patch fixed the issue.
Akemi
Keith Keller wrote:
I was looking at this bug which my machines are currently experiencing:
https://bugzilla.redhat.com/show_bug.cgi?id=883905
The proposed patch is literally one new line in the XFS codebase. So since the patch is so straightforward, I had a crazy idea that I would build my own kernel with this patch, and test it out to see if it worked. (It's been many years since I built my own kernel, so that would be an adventure in and of itself.)
Looks like this problem will be fix in kernel-2.6.32-358.5.1.el6 - see https://bugzilla.redhat.com/show_bug.cgi?id=921958
James Pearson
-
Akemi Yagi -
James Pearson -
Keith Keller