[Linux-cluster] dlm_sendd 100% CPU

[David Ayre]

Someone reported earlier a similar problem with dlm_sendd taking up  
100%:

http://www.mail-archive.com/linux-cluster@redhat.com/msg01002.html

Could our problem be with bugs in our kernel and GFS versions ? (2.6.9- 
X) ?



> Any ideas why dlm_sendd is taking up 100% CPU ?
>
> We have this intermittent problem which seems to happen when a group  
> of users log into our servers at the same time, possibly opening/ 
> writing files from the same mount point.
>
> Should dlm_sendd ever take up 100% of the CPU ?
>
> Any help would be greatly appreciated.... i mean greatly.... like  
> sell my soul "greatly".  If there is any additional diagnostic info  
> i can provide, please let me know.    /var/log/messages just shows  
> failed SMB pipes/write failures due to the high load/lockups.
>
> Version info:
> ----------------------------
> Kernel:   2.6.9-55.0.6.ELsmp #1 SMP
> GFS:  2.6.9-72.2.0.7
> SMB: 3.0.10-1.4E.12.2
>
> Top:
> ----------------------------
> top - 15:48:44 up 53 days, 13:35,  3 users,  load average: 57.40,  
> 48.47, 37.00
> Tasks: 257 total,  64 running, 193 sleeping,   0 stopped,   0 zombie
> Cpu0  :  0.3% us, 99.7% sy,  0.0% ni,  0.0% id,  0.0% wa,  0.0% hi,   
> 0.0% si
> Cpu1  :  0.0% us, 100.0% sy,  0.0% ni,  0.0% id,  0.0% wa,  0.0%  
> hi,  0.0% si
> Cpu2  :  0.3% us, 99.7% sy,  0.0% ni,  0.0% id,  0.0% wa,  0.0% hi,   
> 0.0% si
> Cpu3  :  0.0% us, 100.0% sy,  0.0% ni,  0.0% id,  0.0% wa,  0.0%  
> hi,  0.0% si
> Mem:   4147164k total,  1089108k used,  3058056k free,     5200k  
> buffers
> Swap:  2031608k total,      224k used,  2031384k free,   795340k  
> cached
>
>  PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND
> 4218 root       9 -10     0    0    0 R  100  0.0   1066:10 dlm_sendd
> ...