CPU load spike

[Yan Wong]

Hi folks,

My Unix box has just reached a CPU average workload to 3.xx and then back to 1.xx as normal.

In the back of ps -ef, I find there are 50 prcesses as below:
root 61 0 0 Feb 10 ? 0:00 vx_inactive_thread


any ideas about what is happening?

Yan

[Patrick Wallek]

It sounds like you just had a spike of work for the system to do and when it finished you load average went back down. It's not anything to worry about.

The vx_inactive_thread processes are a result of VxFS (JFS) file systems. If you have VxFS file systems configured, then there is nothing you can do about those processes. Don't worry too much about them. They shouldn't use much, if any, CPU.

[Rajeev Shukla]

Hi Yan,
These are system process created by vxfs(JFS) filesystem you should not worry about. The spikes in the CPU load are normal as long as they dont last for a while. If the load remains high for some time then its a concern which then can be checked by top command.

Cheers
rajeev

[Michael Tully]

The run-level of servers will always fluctuate. The reason is that jobs wait to be processed and then are run. If there are a number to be processed, then the queue is greater, less jobs, queue is less.
The best way to monitor these fluctuations is to monitor the performance of your systems. There are a couple of ways. You can implement 'sar' so that it collects information daily into data files, or have the measureware agent of the glance product collecting data. (glance is not free, but their is a 60 day trial licensed copy on your Application CD set.

[Yan Wong]

When I grep this inactive_thread in other 2 Unix box with UX v11.0, I find 1 only. Is v10.20 and v11.0 making this different?

[Patrick Wallek]

Yes, 10.20 vs. 11.0 could very well make the difference. I've got a BUNCH of vx_inactive_threads on my 10.20 box, but none on a couple of 11.0 boxes.

As I said previously, there is nothing to worry about from these processes on your 10.20 box.