Hang after reboot

[Guillou_2]

Hi,

We have an alphaserver 1000 4/200 with VMS6.2 and shadowing. Sometimes after few minutes after the reboot the system hangs new connexion is impossible even from the console system.

The last time, just before losing the connexion, a sho dev d showed the dsa0 with status "MountVerify mounted", we halted the server and we used the console command crash to get a dump.

A show summary on this dump shows some interactive process in RWMPB state and some process with username system in COMO.

Any idea about our problem?

Thanks in advance for your help

cdt

Steph

[Ian Miller.]

Is DSA0 the system disk?

(RWMPB - Resource Wait Modified Pagewriter Busy)

I would suspect this is due to the system disk not responding to I/O and there being a pagefile on the system disk.

Anything in the errorlog?

[Guillou_2]

Hi Ian,

yes it's the system unit where the pagefile and the swapfile are.

Unfortunately there is nothing in the errorlog, during normal activity this unit doesn't take error and this problem doesn't occur during all reboot...

[Jan van den Ende]

Well,

You will get SOME reaction after MVTIMEOUT seconds.
The default would be 3600, also one hour. If you have that much patience, it would be VERY interesting what happens then.

as a TEMPORARY measure you could modify MVTIMEOUT to, say. 300, being 5 minutes.
So, after you crashed the system to get the dump, BOOT -FLAG=0,1 (assuming booting from SYS0, but I assume you will know and understand if otherwise), and is SYSBOOT> you do a SET MVTIMEOUT 300
and CONT

I'm interested in the outcome, and maybe we will can take it from there.

btw, if you report back, the values of the various SHA*TMO* parameter would also be interesting

Jan

[Volker Halle]

Steph,

do you have a console terminal ? OPCOM and BROADCAST enabled ? Have there been any messages issued to OPA0: immediately before the hang ?

You need to analyze the forced crash to try and find the reason for the DSA0: apparently not doing any IOs. Start with
SDA> SHOW DEV DSA0

What's the status of the disk and it's members ? Any non-zero error count (also check the PKx0 SCSI adapters). Any IOs pending in the device or port queues ?

Volker.

[Volker Halle]

... and if it's a problem with just one member, try decreasing SHADOW_MBR_TMO and wait what happens if this time is exceeded.

Decreasing MVTIMEOUT won't help much, as all pending IOs to DSA0: will be aborted and returned with VOLINV once MVTIMEOUT has been exceeded. And then what is the system supposed to do with an inoperable system disk ?

Volker.

[Ian Miller.]

the unsaved error log buffers in the crash dump may contain interesting entries. Extract them with SDA command CLUE ERRLOG and analyse the resultant file with ANAL/ERROR, DIAG or other error log analysis tool of your choice.

[Cass Witkowski]

Did you happen to lose some memory? the RWMPB and processes on COMO usually mean low memory. Perhaps some of your memory failed.

Cass

[Guillou_2]

Thanks all for your help

Unfortunatly i didn't had the good behaviour to get all the necessary informations to diagnoze the problem

I didn't save the dump but i extract some informations from it with (read/exec,show crash, show summ,sho stackclue stack). I realize that i would have to execute a clue errlog and a sho device to get more interesting information...

There is nothing of interest on OPA0: just before the hang and no error on the two members of dsa0

I must get more information, I will try to save the dump if this problem comes back

Steph

[Wim Van den Wyngaert]

Had the same problem today with a Alphaserver 1000 and VMS 7.3. Boot was near its end (decwindows) and system blocked. No protocol allowed to login, console was no longer reacting, no messages. Had to power cycle. Now everything is up again.

Wim