ERRORE CODE : HALT CODE 5

[Anas Qattea]

Hello,

Since last four weeks I have noticed on the ALPHA server AS4100/OpenVMS V6.2 it does reboot by itself (once every 4-5 days) with HALT CODE 5.

No configuration changes have been made to the system and no application installed recently .

Suggestions are very welcome.

[labadie_1]

Hello

Welcome to the OpenVMS forum

May be you have some data in the error log file, a
$ dir/dat/siz sys$errorlog:errlo*.*
should show several errlog.sys / errlog.old

and a command like the following, for reading the last 28 days

$ ana/err/since="-28-0"/include=bug sys$errorlog:errlog.sys
(or the corresponding period errlog.old;xxx)

and

$ ana/err/since="-28-0"/sum sys$errorlog:errlog.sys
may be useful too
(or the corresponding period errlog.old;xxx)

may show interesting data.

[Wim Van den Wyngaert]

Check the new files in sys$common:[syserr].
Not sure if 6.2 made them but try.

Wim

[Volker Halle]

Anas,

a HALT CODE 5 is a KERNEL MODE HALT.

The system has executed a HALT instruction in kernel mode. This does not normally happen in OpenVMS, so it's likely, that the PC value has been corrupted and is pointing into random data.

Make sure that the console environment variable AUTO_ACTION is set to RESTART, then you will get a HALT crash written and you can look into the crashdump for further analysis.

Volker.

[Anas Qattea]

Thanks Labadie Vim and Volker
$ ana/err/since="-28-0"/sum sys$errorlog:errlog.sys ...It shows me FATAL BUGCHECK in Kernel Mode.

[Anas Qattea]

Volker;
Do you think there is a hardware problem with one of the installed CPUs (2x533 MHZ) .
This AS4100 Alpha is a cluster member.
Any suggetiion to solve this problem ?

[Volker Halle]

Anas,

if you have a valid dumpfile on your system disk, the CLUE utility would provide a basic crash footprint during startup.

Try $ TYPE CLUE$HISTORY, it should show a 1-line summary of the most recent crashes. Consider to attach the file pointed to by the CLUE$HISTORY logical to your next reply.

For more details, you would also have to supply the CLUE files from each crash - as found in CLUE$COLLECT:CLUE$node_ddmmyy_hhmm.LIS

The minimum required information would be the output of the bugcheck from ERRLOG.

Volker.