[Info-vax] system-f-vasfull problem on alpha vms 7.1

[Bob Gezelter]

On Mar 10, 4:21 am, Chris Casey <chris.ca... at NTLWorld.com> wrote:
> On Mar 9, 6:41 pm, Volker Halle <volker_ha... at hotmail.com> wrote:
>
> > Chris,
>
> > if 'everything appears to hang', can you do SHOW SYSTEM ? Any process
> > in an unsual state (RWxxx) ? Does MONI SYST/INT=1 get updated during
> > that hang period ? Does it show any high resource usage ?
>
> > It could still be nonpaged pool. Trying to get a large packet from the
> > variable list, which is very fragmented, would take some time and
> > affect the overall system. But there is not much you can do in V7.1 to
> > diagnose this.
>
> > Make sure to increase NPAGEDYN to at least 16 million for the next
> > boot.
>
> > Volker.
>
> Volker,
>
> the monitor freezes as well. Also terminal echo stops. The system is
> totally unresponsive.
>
> Chris

Chris,

There are two possibilities:

- examine the running system to determine what is happening BEFORE the
system gets into a degraded state
- get a crash dump of the system when it "freezes" (i.e., becomes
unresponsive)

Unresponsive is a perception, not a causal diagnosis. A perceived
freeze can be caused by a number of problems, including:

- thrashing
- high paging combined with a fragmented page file
- run away user process at a high priority
- system bug
- IO problem

Admittedly, here in the newsgroup, we are limited by a very narrow
window. A good procedural start would be to ensure that when this next
happens, that a system dump is actually taken as part of the rebooting
sequence. At that point, the dump file can be taken offline and
reviewed in depth to determine what the state of the system was at the
point of the failure.

- Bob Gezelter, http://www.rlgsc.com