Oninit cdi / Archive
Talk to an expert →

Removing semaphore shared memory segment aborts other informixservers

Posted in 2010

Topics: Backup & Restore, Error Codes & Troubleshooting

red_valsen — — source: Usenet: comp.databases.informix

Due to an interrupted ontape restore, I was left with a down informixserver with its shared memory segments still being reported by ipcs. I carefully removed all the segments using ipcrm, being careful not to remove any used by the other two instances on the host, but when I removed a semaphore segment from the fouled instance, the other two abruptly crashed with "Assert Failed: semop: errno = 36" messages appearing in the Informix message logs for each. Has anybody else experienced this? I can find no mention of this behavior in Informix manuals, online, etc. I assume that in some way the other instances are using that segment of memory. Using IDS 11.50.FC4 on HPUX 11.23 (memory management with this O/S has always seemed problematic).

↪ replying to red_valsen
Doug Lawry — — source: Usenet: comp.databases.informix

See the new "onclean" command which does the equivalent of "ipcrm" for you but more safely! Regards, Doug Lawry

↪ replying to red_valsen
Fernando Nunes — — source: Usenet: comp.databases.informix

On 14-12-2010 13:41, red_valsen wrote: > Due to an interrupted ontape restore, I was left with a down > informixserver with its shared memory segments still being reported by > ipcs. I carefully removed all the segments using ipcrm, being careful > not to remove any used by the other two instances on the host, but > when I removed a semaphore segment from the fouled instance, the other > two abruptly crashed with "Assert Failed: semop: errno = 36" messages > appearing in the Informix message logs for each. > > Has anybody else experienced this? I can find no mention of this > behavior in Informix manuals, online, etc. I assume that in some way > the other instances are using that segment of memory. > > Using IDS 11.50.FC4 on HPUX 11.23 (memory management with this O/S has > always seemed problematic). errno 36 is "removed identifier". I never experienced this. It's hard not to think that you removed the wrong set... On the other hand you say you had other two instance crashing... It would be easier to accept a mistake if it only affected one... This would never be mentioned in the manuals, because if anything it would be a bug... Please use onclean as already suggested. Regards.

Glossary