[Fedora-directory-users] db2ldif causes server to hang

[Steve Halstead]

Hi,

We are currently having problems where by a Fedora DS LDAP server which 
has been running for several days begins to hang. The consequence of 
this is that further LDAP commands (e.g. ldapsearch and db2ldif) 
triggered by cron jobs start but fail to complete. The only way to 
recover is to run "killall -9 ns-slapd" and then restarting the server 
in the usual way.

On an hourly basis, db2ldif is used to export our LDAP repository to an 
LDIF file. Looking through the logs, it would appear that the server can 
start to hang if db2ldif is interrupted in some way.

The last time that we saw the server hang, in the access log we had 
ldapsearch and "Netscape Replication Start Session" both accessing the 
server at the same time as our db2ldif export process was running. In 
the error log it could be seen that db2ldif stopped mid-way through 
exporting users. All further executions of db2ldif failed to complete.

It would seem that db2lif is generating some sort of lock which isn't 
released if it is interrupted.

I have managed to reproduce this server hang by running db2ldif and 
killing it with CTRL+C.

We are currently running fedora-ds 1.0.2 on RHEL 4 but I have tried 
1.0.4 and had a similar experience.

Has anybody else had a similar experience, or even better, know how to 
fix it?

Thanks,
Steve