On 6/7/2016 6:31 PM, Johnny Hughes wrote: > On 06/07/2016 03:18 PM, Wayne Betts wrote: >> On 6/7/2016 11:22 AM, Johnny Hughes wrote: >>> On 06/07/2016 09:56 AM, Wayne Betts wrote: >>>> I'm not sure if this is an appropriate place for this question - please gently >>>> steer me to a better place if you think there is one. >>>> >>>> I have a machine using a kernel package from the CentOS Community Build System >>>> (specifically kernel-3.18.21-16.el6 from >>>> http://cbs.centos.org/koji/buildinfo?buildID=4757, but my question applies to >>>> other CBS kernel versions as well from what I can tell) >>>> >>>> The machine is crashing with some regularity on Sunday mornings. Looking at a >>>> kdump core dump might help narrow down the cause of the crashes. >>>> >>>> But the CBS kernel packages don't appear to include kernel-debuginfo packages, >>>> and the kernel.spec file appears to me (a non-expert for sure) to have had the >>>> necessary pieces removed, so I don't know how to rebuild from the srpm and get >>>> the debuginfo packages that way either. >>>> >>>> If someone could help me find or build the kernel-debuginfo package, it will be >>>> appreciated. >>>> >>> >>> Those kernel packages do not build debuginfo files. That is also not >>> nearly the newest package. There is a 3.18.34-20 package here: >>> >>> http://cbs.centos.org/koji/buildinfo?buildID=11179 >>> >>> You can see the packages that are built and look at all the build logs >>> and root logs, etc, by going to: >>> >>> http://cbs.centos.org/koji/ >>> >>> and adding kernel in the search box .. then picking the version you are >>> interested in. >>> >>> But, as you can see in the above link, there are no debuginfo files >>> produced. >>> >>> However, we do take updates directly from the kernel.org 3.18 LTS kernel >>> tree, so there are a lot of things that have been fixed (and a lot of >>> security updates) since 3.18.21 .. in fact you can see all of the >>> changelogs starting with : >>> >>> https://cdn.kernel.org/pub/linux/kernel/v3.x/ChangeLog-3.18.22 >>> >>> and using 23, 24. 25, 26, 27, 28, 29, 30, 31, 32, 33, and 34 for the >>> last number to review all the changelogs. >>> >>> Thanks, >>> Johnny Hughes >>> >> >> >> Thank you for your reply Johnny. I tried the latest, >> kernel-3.18.34-20.el6.x86_64, and that backfired - attempting to read from our >> ceph cluster triggers an immediate crash. >> >> The behaviour of crashing on cephFS access starts at 3.18.30-20 in the CBS kernels. >> >> Meanwhile, 3.18.25-20 crashes early at boot, failing to mount the root >> filesystem (so I don't know if the Ceph crash happens at this kernel version). >> >> I'm leaving the system live now at 3.18.25-19. Assuming this hasn't introduced >> some additional problem(s), it will be a while before we know if this has >> reduced or eliminated the crashes we are experiencing. > > Also, are you using this with XEN as well or just using this because it > is a newer kernel? > We are not using this with Xen. The Ceph project recommendation is to upgrade Ceph clients from the "stock" 2.6.32 Scientific Linux/RHEL 6 kernels (see the notes at http://docs.ceph.com/docs/master/start/os-recommendations/ ). We used some 3.10 LTS series kernels for a bit, but our Ceph expert subsequently suggested the use of the CBS 3.18 series kernel for our CephFS clients, though I don't know if there was a specific reason other than 18 > 10. (I'd guess he found some anecdotal evidence or a recommendation online from a fellow CephFS user to use the CBS 3.18 kernels instead of the LTS 3.10 kernels for Ceph, but don't really know.) -Wayne