 |
2 correction: Currently used kernel version - 3.10.0-229.1.2.el7.x86_64. Core number - for now only CPU0 is affected (network queues are distributed between different cores). |
 |
Probably related to this known issue: https://access.redhat.com/solutions/1409393 |
|
|
See also: https://access.redhat.com/articles/879293 |
|
|
@toracat, unfortunately, Red Hat KB is accessible only by subscribers. Nevertheless, it seems that this issue has been fixed by the following commit: https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/drivers/dma/ioat?id=da87ca4d4ca101f177fffd84f1f0a5e4c0343557 More info: * https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1291113 * https://lwn.net/Articles/588457/ I've faced the same issue and going to try recompiled CentOS 7 kernel with abovementioned patch included. |
|
|
@post-factum, Thanks for the links. Please do let us know of the result of patching. |
|
|
Now that the referenced patch is in the mainline stable kernel from kernel.org, one can test-install kernel-ml [1] from ELRepo and see if the problem does not occur there. Can someone test this? [1] http://elrepo.org/tiki/kernel-ml |
|
|
A good piece of info from RH: "Yes, the patch above is the expected solution. It's also scheduled to be included on RHEL 7 kernel." If the patch is not in the next el7 kernel update, I will attempt to include it in the centosplus kernel. |
|
|
@toracat, the problem is that to raise up the issue one need some time. First occurrence happened to us after 3 months of uptime, second one — just after 8 days. We've scheduled our production server for kernel maintenance on Sunday night. |
|
|
kernel-3.10.0-229.11.1.el7 is out. The plus kernel now has the patch quoted in this bug report. |
|
|
Sadly, 3.10.0-229.11.1.el7 does not have this patch. Anyway, I've recompiled 229.7 with it, and our server runs for 10 days with no problem. Should still monitor, though. |
|
|
@post-factum, Right, the distro kernel still does not have this patch. Good to know the patch seems to be working on your system (so far). If you switch to kernel-plus-3.10.0-229.11.1.el7.centos.plus, it should work like the one you compiled. |
|
|
OK, server survived ~25 days of successful uptime with self-compiled kernel. I had to reinstall it due to disk replace, now switched to kernel-plus. |
|
|
The distro kernel 3.10.0-229.14.1.el7 has the patch. It is also confirmed that the RHEL 7.2 beta kernel (-306) also has it. |
- Anonymous
