CentOS 5.5 : openLDAP server causes server to hang

Hi All,

I'm having some issues (https://www.virtualmin.com/node/16904) which are causing my server to start to kill jobs out of memory. One of those jobs is the LDAP server. Here's where there's a nasty chain reaction and upon reboot the system hangs before SSHd or the serial console are up. Only solution is to boot a recovery CD (or use Xen dom0 if it's a VPS) and mount the hard drive to fix things.

Here's what shows up on the console (for future googlers)

Starting HAL daemon: [ OK ] Checking configuration files for slapd: bdb_db_open: unclean shutdown detected; attempting recovery. bdb_db_open: Recovery skipped in read-only mode. Run manual recovery if errors are encountered.

And then the machine halts there and you're SOL.

From this fine blog : http://techarold.blogspot.com/2006/07/more-openldap-recovery.html the solution is to run

/usr/sbin/slapd_db_recover -v -h /var/lib/ldap Finding last valid log LSN: file: 1 offset 157293 Recovery starting from [1][156143] Recovery complete at Mon Jan 17 10:02:19 2011 Maximum transaction ID 80000056 Recovery checkpoint [1][158343]

And then you can start the LDAP server.

However in order to get that far I had to mount the hard drive from a recover CD and then prevent the LDAP server from starting :

rm /etc/init.d/rc3.d/Sldap

And then I could boot the machine to run the above fix command. Don't forget to put the LDAP server back in.

So for virtualmin I would strongly suggest an additional LDAP configuration command to prevent this deadlock. Here is the upstream bug : https://partner-bugzilla.redhat.com/show_bug.cgi?id=613966

To fix this I added -u to the 'slaptest' command at all times to make sure it would never hang. Edit /etc/init.d/ldap and change the configtest() function to match the following :

function configtest() { local user= ldapuid= dbdir= file= # Check for simple-but-common errors. user=ldap prog=basename ${slapd} ldapuid=id -u $user # Unaccessible database files. slaptestflags=-u

I changed the last line only. IF you do not do this, you may find your server hung and unresponsive w/out SSH or Console available.

-m