[Date Prev][Date Next] [Chronological] [Thread] [Top]

Re: Weird problem -- seeking insight



--On Friday, January 16, 2004 12:03 PM -0500 Frank Swasey <Frank.Swasey@uvm.edu> wrote:

Folks,
  I have four ldap servers.  They are all running RedHat 9, Openldap
2.1.22 with BDB 4.1.25 patch 1 (I built the openldap rpm myself).  We
have one Master (R/W) and three slaves (R/O) servers.  The problem I'm
having is that during the nightly update process (where the
authoritative sources of affiliations are merged and the existing
database is updated) of a large number of modrdn and modify operations,
one of the three slaves has been failing with deadlock problems for the
past three days while the other two slaves and the master are doing just
fine.

  The specific error message issued is:

Jan 15 02:27:45 blorp slapd[7830]: => bdb_idl_insert_key: c_put id
failed: DB_LOCK_DEADLOCK: Locker killed to resolve a deadlock (-30996)

  I haven't been able to find in the documentation for SleepyCat (I
haven't gone searching in the source code yet) what that is trying to
tell me...  I'm looking for anyone who can give me an idea of what the
problem is, whether it could be software or is definitely hardware
related (as I said this is one out of three R/O machines being fed by
slurpd from the R/W).

Thanks for any information. 

Hi Frank,

Have you tried a db_recover on that system when slapd is stopped? Also, have you tried reloading that slave from the master, and seeing if that resolves the problem the next nightly run?

--Quanah

--
Quanah Gibson-Mount
Principal Software Developer
ITSS/TSS/Computing Systems
ITSS/TSS/Infrastructure Operations
Stanford University
GnuPG Public Key: http://www.stanford.edu/~quanah/pgp.html