[Ocfs2-users] OCFS2 KVM Crashes Yet Again !

[netbsd at tango.lu]

Hello,

Find the full log below:

https://urldefense.proofpoint.com/v2/url?u=https-3A__paste.ubuntu.com_25625787_&d=DwICAg&c=RoP1YumCXCgaWHvlZYR8PQcxBKCX5YTpkKY057SbK10&r=QxGl6UoyzTJm_1fAz5ZR9izvWJhWcqbtYn-0afBpa7A&m=5BOEZ6shfWftzp2R4SRZiOEZXmUvTYJZyua8Tuqlya4&s=Wj5bNrDYurcyqvciOCmJAZyQjrpSQjQVP_9-laqPiso&e= 

VM was restarted at 9:27 and no problem since then. We are rsyncing 
about 2TB data (a lot of small files) between 2 OCFS shares on the same 
vm:


/dev/vdc                      4.8T  2.8T  2.1T  58% /mnt/s1
/dev/vdf                      4.8T  985G  3.9T  21% /mnt/s2

rsync -av --numeric-ids --delete /mnt/s1/ /mnt/s2/


On 2017-09-27 10:53, Gang He wrote:
> Hello netbsd,
> 
> The ocfs2 project is still be developed by us (from SUE, Huawei,
> Oracle and H3C. etc.).
> If you encountered some problem, please send the mail to ocfs2-devel
> mail list, we usually watch that mail for ocfs2 kernel related issues.
> 
> 
> 
> 
>>>> 
>> Hello All,
>> 
>> I wrote earlier about our OCFS2 crash issue in KVM due to bug in the 
>> SMP
>> code.
>> 
>> For this we come up with a solution:
>> 
>> Instead of using multiple vcpus
>>    <vcpu placement='static'>8</vcpu>
>> 
>> using a single one and multiple cores instead:
>>      <topology sockets='8' cores='8' threads='1'/>
>> 
>> And applying key tune options to sysctl.conf:
>> 
>> vm.min_free_kbytes=131072
>> vm.zone_reclaim_mode=1
>> 
>> Seemed to be helped, the fs did not crash right away when we were
>> hammering it with apache benchmarks with 10000 requests however last
>> night I started a large rsync operation from a 5TB OCFS2 FS mounted in
>> the VM to another OCFS2 mounted in the same VM and ended up with:
>> 
>> https://urldefense.proofpoint.com/v2/url?u=https-3A__ibb.co_gFeGg5&d=DwICAg&c=R
>> oP1YumCXCgaWHvlZYR8PQcxBKCX5YTpkKY057SbK10&r=QxGl6UoyzTJm_1fAz5ZR9izvWJhWcqbtY
>> n-0afBpa7A&m=cYprGRHz-oQmhnx4HIke8sTdCG_tf8Jb-rF6sHnYLnk&s=ajWfQIlUZOpElFWxoKcmvTI
>> k7J3PpuCJITcnXfJQHrc&e=
> From the kernel crash backtrace, this problem should be that long time
> to acquiring spin_lock triggers a NMI interruption.
> Could you give a detailed reproduce steps? since we want to reproduce
> this issue in local, then try to fix it.
> 
> 
> Thanks
> Gang
> 
>> 
>> After trying a lot of different kernels starting from the 3.x series,
>> now we are using 4.13.2 latest kernel with default configuration but
>> these issues still present. Is this OCFS2 project still being 
>> developed?
>> With this crashing and unreliability it cannot be used in production
>> unless you put in place bunch of safeguards to reset out the whole
>> virtualmachine when it crashes.
>> 
>> Thanks
>> 
>> _______________________________________________
>> Ocfs2-users mailing list
>> Ocfs2-users at oss.oracle.com
>> https://oss.oracle.com/mailman/listinfo/ocfs2-users