LinuxQuestions.org - Showing strange load average, after reboot

- Solaris / OpenSolaris (https://www.linuxquestions.org/questions/solaris-opensolaris-20/)

- - Showing strange load average, after reboot (https://www.linuxquestions.org/questions/solaris-opensolaris-20/showing-strange-load-average-after-reboot-4175634187/)

Showing strange load average, after reboot

Hello,
This is x86, Solaris-10 server. Due to some reason, this was extremely slow and I had to reboot it. It came back online now and working fine. After reboot, it shows me strange load average value. Though with time, it is coming down. Can somebody explain it, why it is so ?

Code:

Before reboot -

bash-3.2# w

 11:22am  up 221 day(s), 22:27,  2 users,  load average: 0.14, 0.15, 0.44

User    tty          login@  idle  JCPU  PCPU  what

oracle  pts/2        11:04am    9                -ksh

matt  pts/3        11:11am                      w

bash-3.2# 



After reboot -

bash-3.2# w

 11:21am  up 8 min(s),  2 users,  load average: 134.27, 23810.79, 24400.47

User    tty          login@  idle  JCPU  PCPU  what

oracle  pts/1        11:13am    8                -ksh

matt  pts/3        11:16am                      w

bash-3.2# w

 11:23am  up 10 min(s),  2 users,  load average: 21.56, 16505.54, 21597.75

User    tty          login@  idle  JCPU  PCPU  what

oracle  pts/1        11:13am    10                -ksh

matt  pts/3        11:16am                      w

bash-3.2# w

 11:28am  up 15 min(s),  2 users,  load average: 0.44, 6636.28, 15945.94

User    tty          login@  idle  JCPU  PCPU  what

oracle  pts/1        11:13am    14                -ksh

matt  pts/3        11:16am                      w

bash-3.2#

The load average is a decayed average of the time spent by threads either running on or waiting for a CPU. The last value displayed is taking into account the values observed during a longer period of time (~15 minutes) so is staying high for a longer time while the first one is taking into account values from the last minute or so, so is reaching reasonable values faster.

It seems there are tens or maybe hundreds of thousands threads waiting on your machine early in the boot phase, which is not a common situation.

To investigate further, you can reboot, figure out how many threads are waiting on the run queue with the "vmstat" command and identify processes with a large number of threads with the "prstat -m" command.

Is there any way to check, without reboot ?

No. You need to have these statistics already collected and stored.

Okay... Let me check, if I can take downtime or will wait, till it slows down again and come in same situation.
Thanks for explaining this

Note that the probability for the high load average at boot and the poor performance before reboot to be related is low.