Re: [Moosefs-users] mfsmetalogger segfault

[Robert S. <rsa...@ne...>]

I saw similar crashes when mfsmaster was unresponsive for long times. 
Specifically around the hour when it dumps the meta-data to disk and the 
master becomes very busy.

If this sounds familiar ensure that you have 2 times the memory usage of 
mfsmaster available on the master and it should be better. This fixed it 
for me.

Robert

On 9/26/11 1:57 PM, Ricardo J. Barberis wrote:
> El Lunes 26 Septiembre 2011, Laurent Wandrebeck escribió:
>> Hi,
>>
>> CentOS 6 x86_64, mfs 1.6.20 package from rpmforge.
>> mfsmetalogger[2043]: segfault at 7f261ddf6380 ip 00007f261dae42e5 sp
>> 00007fff6dee7d50 error 7 in libc-2.12.so[7f261da7e000+175000]
>>
>> Has anyone seen that ?
> Under CentOS 5 x86_64, mfs-1.6.20 from rpmforge, I got this on September 18:
>
> kernel: mfsmetalogger[1967]: segfault at 0000000000000060 rip 0000003fb4a612ed
> rsp 00007fff709d0860 error 4
>
> On another metalogger server, from September 8:
>
> kernel: mfsmetalogger[2150]: segfault at 0000000000000008 rip 00002b2f88718350
> rsp 00007fff9c3044e0 error 4
>
> kernel: mfsmetalogger[2149]: segfault at 0000000000000008 rip 00002b16ea2f9350
> rsp 00007fff58b948d0 error 4
>
>
> I think I can reproduce it (thought I haven't really tested it) by unpluggin
> the network cable, or at least I've seen it happen a couple of times while
> moving some servers to a new switch.
>
> It's an easy test so I guess I could try it and see if it crashes.
>
> Cheers,