I was checking the output of top today and noticed a load average that I would've thought should be associated with a pretty unresponsive system.
Code:
top - 13:40:39 up 145 days, 1:19, 3 users, load average: 11.00, 11.00, 11.00
Tasks: 293 total, 1 running, 292 sleeping, 0 stopped, 0 zombie
Cpu(s): 0.3% user, 1.6% system, 0.0% nice, 98.0% idle
Mem: 904496k total, 897572k used, 6924k free, 205308k buffers
Swap: 506008k total, 112476k used, 393532k free, 180396k cached
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
16921 root 16 0 1216 1216 840 R 1.6 0.1 0:00.26 top
13105 client 9 0 2184 2184 1644 S 0.3 0.2 0:05.46 sshd
1 root 8 0 72 64 44 S 0.0 0.0 0:27.72 init
2 root 9 0 0 0 0 S 0.0 0.0 0:00.40 keventd
The system seems to perform just fine despite the 'load' but I learned everything I wanted to about the Unix load average metric:
http://www.teamquest.com/resources/g...ay/5/index.htm
including that processes in a disk sleep state contribute to the run queue length but not necessarily significantly to CPU usage.
The following processes are in the disk sleep state:
Code:
karl 27500 0.0 0.2 28380 2432 ? D Jul27 0:05 kdeinit: kded
root 17604 0.0 0.0 1632 744 ? D Aug24 0:04 /usr/bin/updatedb -c
root 30779 0.0 0.0 1632 744 ? D Aug25 0:04 /usr/bin/updatedb -c
nobody 1411 0.0 0.2 7348 2012 ? D Aug26 0:00 /usr/sbin/smbd -D
nobody 1414 0.0 0.1 7348 1688 ? D Aug26 0:00 /usr/sbin/smbd -D
nobody 1416 0.0 0.1 7348 1688 ? D Aug26 0:00 /usr/sbin/smbd -D
root 5411 0.0 0.0 1632 744 ? D Aug26 0:03 /usr/bin/updatedb -c
root 20449 0.0 0.0 1632 744 ? D Aug27 0:03 /usr/bin/updatedb -c
root 24443 0.0 0.0 1632 744 ? D Aug28 0:04 /usr/bin/updatedb -c
root 20044 0.0 0.1 2712 1676 ? D Aug29 0:00 bash
root 24758 0.0 0.0 1632 744 ? D Aug29 0:03 /usr/bin/updatedb -c
I'd hoped that the root of the problem here was a smbmount that was disconnected in an ugly fashion so I unmounted ("umount -fl /mnt/sharename") the samba mounts and killed the process that was a parent to the smbd processes. The result was only that init is now the PPID of the smbd processes. I'm assuming that the updatedb processes are just a result of the search iterating over the broken samba mount.
How can I find what these processes are waiting for? I don't think they're going to timeout (most are nearly a month old)
This is a Slackware 10.2 box.
cat /proc/version:
Linux version 2.4.31 (root@tree) (gcc version 3.3.5) #6 Sun Jun 5 19:04:47 PDT 2005