Online Expert Day - HPE Data Storage - Live Now
April 24/25 - Online Expert Day - HPE Data Storage - Live Now
Read more
Operating System - OpenVMS
cancel
Showing results for 
Search instead for 
Did you mean: 

DECW$SERVER computable

Andrea Angelo
Occasional Visitor

DECW$SERVER computable

We have two Itanium rx2620 with OVMS 8.2-1. We sometimes find both of them blocked: no access from the console is possible. With a network login (set host or telnet) we have found CPU load al 100% and DECW$SERVER computable. The uptime seems to be one month. The console is the unique graphic port on the management card and it is used only few hours a week with DECterm. DECW$STARTUP RESTART is not useful and seems to get a worse situation (access via network is no more possible).
1 REPLY
Hoff
Honored Contributor

Re: DECW$SERVER computable

That's likely a bug. Somewhere.

Check the firmware revisions for the various boards involved -- MP, etc -- and check the patch level for OpenVMS, and then ring up HP support and ask.

There have been MP firmware revisions for various Integrity servers, and there have also been mandatory OpenVMS ECOs and ECOs for various DECwindows components over the years. Any of which could conceivably remediate a problem (somewhere) that triggers a loop (somewhere else) that causes some piece of DECwindows to go computable.

PCSI and UPDATE ECOs would be a start; the master ECO list and the released ECO kits are available at ftp://ftp.itrc.hp.com/openvms_patches/i64/V8.2-1/

After the mandatory and the relevant ECOs, the next step would involve doing some PC tracing, to figure out where the looping is occurring within the particular components, and working backwards from there to sort out why this is arising. (This is where HP services generally gets involved.)

I'd probably apply ECOs and any new firmware when scheduling permitted, leave the box to loop again, then -- if the loop arises anew -- try to use SHOW PROCESS/CONTINUOUS to capture some PC addresses in the looping process (to get an idea where in the code the loop is occurring), and I'd then force a crash and write out a crashdump for HP to look at.

As a potential work-around pending ECOs and any response from HP, disable DECwindows (there's a logical name used for this in SYLOGICALS.COM or such, DECW$IGNORE_WORKSTATION) and use a serial or MP connection into the box.