We've been having intermittent problems with backups, much like the ones I
see previously described in this list. (Reference:
http://www.ornl.gov/its/archives/mailing-lists/alpha-osf-managers/1994/10/msg00269.html)
The original problem:
dump: Mapping (Pass II) [directories] (continued)
dump: Cannot get semaphores
get_semaphore(): semget() for fifo_free_semid: No space left on device
dump: SIGTERM received
dump: The ENTIRE dump is aborted
The previous system administrator tried to fix it by writing a script that
uses ipcrm to remove semaphores that look like they were left over from
the last dump before running a new dump. This did not fix the problem
entirely. Now we don't see the "get_semaphore()" errors, but we
occasionally do see:
dump: 1.12% done -- finished in 00:02
dump: Cannot set fifo_free_semid[2]
flusht(): semop(): Identifier removed
dump: Cannot remove semaphores
remove_semaphore(): semctl() for fifo_free_semid: Invalid argument
dump: SIGTERM received
dump: The ENTIRE dump is aborted
I'm not sure what to do now. The previous sysadmin thinks that upgrading
to 4.0E (we're running 4.0D now) might solve the problem. I'm not so sure
it will and would rather not upgrade in the middle of the school year.
My instinct is to try what Harald Mayer & Jan M. Holzer suggested in the
above referenced email to the OSF list and increase the value of the
semmni parameter to 90.
Is my former sysadmin right about upgrading to 4.0E or should I just up
the semmni on 4.0D? If I do up the value of semmni, I'd like to read more
about what it is and how to do it. Jan Holzer recommends the "Guide to
System Administration section 5.4.1.5" but I don't know where that is in
the system or if it's printed, I don't think I have it. In general I've
not rebuilt the kernel before and would love pointers to documentation on
doing it. Thanks,
bjh
Received on Sat Feb 13 1999 - 00:19:51 NZDT