[lug] Re: kupdate and stalls

[Doug Hunt]

Tim:  I don't know anything about kflushd (other than it is some kernel
daemon that deals with memory management) but a google search turned up
this interesting (if not completely informative) tidbit:

http://www.uwsg.iu.edu/hypermail/linux/kernel/9904.2/0468.html

Regards, 

  Doug Hunt

Tim Fredrick wrote:
> 
> Hello,
>    We have a 4-CPU 500MHz Xeon Dell Poweredge here with a bunch (about 256
> accounts and about 32 active logins) here that has started behaving
> badly.  That is, 2 or 3 times in a 30-minute period it will stall for a
> few seconds.  I took a look at "top" and "xosview" and what happens
> is the CPU LOAD goes high up to as much as 20.  Yet we see less activity
> on the individual CPU's, no swap usage, no disk activity, and a lessening
> of process activity.  In terms of process activity, the USR processes
> nearly disappear while the SYS processes diminish in activity.
>    top shows that during this time, we're running "kupdate" and "kupdate"
> 
>     2 root       3   0     0    0     0 DW      0  7.7  0.0  15:40 kflushd
>     3 root       1   0     0    0     0 DW      0  5.8  0.0  34:12 kupdate
> 
> along with an expected mix of user processes.  Can someone tell me what
> these two processes which are part of the kernel do?   Our system's
> set is currently:
> 
>         Redhat 6.1
>         Kernel 2.2.14 SMP
>         2GB RAM (1.8 show up as being used on cat /proc/meminfo)
>         OS and swap space on RAID disk, user files on separate RAID disk
> 
> Can someone advise how I might further figure out what's causing these
> nasty stalls?  Would upgrading the Redhat version or kernel be a way out
> of this?  How about calling Scully and Mulder?
> 
>    Thanks in advance,  --Tim (fredrick at acd.ucar.edu)
>                          NCAR/ACD Computer Systems Administration

-- 
dhunt at ucar.edu
Software Engineer III
UCAR - COSMIC
Tel. (303) 497-2611