High number of blocked processes

[Terry Gibbar]

Is it normal for there to be over 250 processes blocked?

procs
r b w avm free
1 262 0 41001 79563
1 262 0 41001 79521
1 262 0 41001 79521
1 262 0 41001 79521
2 261 0 42425 79521
2 261 0 42425 79513
2 261 0 42425 79579

[Stefan Farrelly]

No, its not normal for that many blocked processes.

you need to do a fuller listing of your processes in order to try to establish what theyre blocked on. Or check sar or glance (if available). Also check for stuck tcp sockets (netstat command).

[Anthony deRito]

No it is not. Try to determine what the processes are blocked on. Use GlancePlus process windows to determine the type of resource the processes are waiting for. Depending on what type of process it is, it can be blocked on buffer cache, disk I/O, priority, inode or many other resources. If you do not have GlancePlus, you will have to use tools like vmstat and sar. Let us know.

Tony

[Terry Gibbar]

There is no WIO:
10:24:06 %usr %sys %wio %idle
10:24:11 7 3 1 90
10:24:16 4 2 1 93
10:24:21 2 1 0 96

Glance shows an extreme amount of processes blocked on OTHER. What would this be telling me?

uop_b_slow 8945 1 155 upds999 0.0/ 0.0 0.0 0.0/ 0.0 5.2mb OTHER

Glance also shows a few processes blocked on MESG.

BBL 1527 1 155 tsbusr 0.0/ 0.0 0.0 0.0/ 0.0 5.1mb MESG

What other commands/utilities would be helpful in determining what the processes are blocked on?

I'm confused. This server appears to have plenty of resources.

[Anthony deRito]

The BBL process I am assuming is the Tuxedo Buliten Board process. Is this true? It is normal for that to spend a lot of time servicing a message event. The fact that the process in question is blocked on "other" really is vague. The help screens say that they are "unknown activities". Take a look at your tables with sar -v and also your VM activity with sar -w. Check also the rate of system calls with sar -c. In fact, in the process wait state window, use the "process system calls" option under "reports". Check out what types of systems calls are being made (if any). This may give a hint.

Tony

[Alan Riggs · Accepted solution]

Check for PHKL_16301 or superseding patches. This patch corrects a known problem in which vmstat incorrectly reports a high number of blocked processes.

[Terry Gibbar]

I do indeed need that patch, but how do I know if the output is truly bogus? Short of applying the patch, of course.

Thanks

[Alan Riggs]

No way to be sure until you apply the patch, but the numbers seem suspiciously high. If your system does not show other symptoms of a serious resource contention, then I think it very likely that the vmstat report is bogus.