RE: Slurpd locks when slave's HD dies (ITS#2862)

[quanah@stanford.edu]

--On Sunday, December 07, 2003 2:58 AM -0800 Howard Chu 
<hyc@highlandsun.com> wrote:

>> -----Original Message-----
>> From: owner-openldap-bugs@OpenLDAP.org
>> [mailto:owner-openldap-bugs@OpenLDAP.org]On Behalf Of quanah@stanford.edu
>
>> I found out rather accidentally today when the HD failed on one
>> of our OpenLDAP
>> servers, that slurpd locked up and would no longer send
>> updates to the replicas.
>>  A normal kill -INT also would not stop slurpd (it had to be
>> done with a kill
>> -9).  After restarting slurpd, updates processed as normal.
>> I imagine some
>> thread lockup happened.  Truss wasn't very useful.
>
> This report is a bit vague. Was the locked up slurpd on the same machine
> that had the failed drive? Or was it on another machine, in the middle of
> updating a hung slave?

The slave was ldap6, and its drive failed.  The master is ldap0, and its 
slurpd locked up, and blocked all further processing of updates to the rest 
of our replica pool.  Interestingly enough, slurpd continued to pull the 
changes from slapd's replog, and push them into slurpd's replog.  Even 
after I rebooted ldap6, slurpd's connection to ldap6 continued to block 
updates to the rest of the replica's.

--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