vxfs_thread jumping to top on linux servers
Updated: 22 May 2010 | 2 comments
I've noticed the vxfs_thread jumping to the top of the process list on linux servers (SLES9/10) We aren't event using vxfs on the servers so it's curious as to why it even shows up as utilizating cpu. There's particular servers where we do use it though so to remain consistent, the same kernel modules and services are started at boot. So it isn't an option to shut it off. Has anyone seen this behavior before or can explain it?
discussion Filed Under:
Comments
RE: vxfs_thread jumping to top on linux servers
RE: vxfs_thread jumping to top on linux servers
we have propably the same issue here.
- Systemload goes up to about 70.
- CPU Usage is low 95% idle time.
- System with Oracle 10g (DB filesystem with rw,delaylog,largefiles,mincache=direct,convosync=direct,ioerror=mwdisable)
If we look for "waiting (not sleeping) threads"
(p.e. ps -eLo uid_hack,pid,ppid,c,stime,tty,time,pcpu,state,wchan,lwp,nlwp,cmd | grep -v " S ") we found a lot of "vx_naio_worker"-threads waiting for I/O
There are 64 threads, all waiting for I/O. Seems to be a hard limit (?)
is it possible to increase/tune the maximum threas?
We are using SLES10 SP1 with Veritas (VRTSvxfs-platform-4.1.40.20-MP4RP2_SLES10)
Would you like to reply?
Login or Register to post your comment.