Open Radar

 
Summary:
Since upgrading to 10.8.x, our large OD sites are experiencing frequent server hangs. Upon investigation it appears that /var/vm is filling up with page files until there is no more space on the drive and the system comes to a complete halt. The only way to recover is via a forced reboot. This often results in a corrupt openldap database which needs to be recovered from backup.
Possible memory leak with slapd based on how much virtual memory it is using, although I am not an expert when it comes to this side of things.

Steps to Reproduce:
Issue only seems to occur at clients that have medium-large populated OD infrastructures. We have dozens of clients using 10.8.x OD for client management only (WGM/MCX Computer/Work groups but no users/auth) which have never experienced this problem. It is only those sites that have hundreds of users in their OD and have client machines authenticating against OD that experience this issue.

Expected Results:
For the server not to fill its boot drive with swapfiles even though there is usually gigabytes of free memory available.

Actual Results:
Boot drive fills, server go bye bye

Version:
10.8.4. Have tried latest 10.8.5 release but issue still present. Although the 10.8.5 delta contains /usr/libexec/slapd the actual version reports the same as the one included in 10.8.4 (slapd -V) so not sure if it's actually changed or not

Notes:
Only workaround at the moment is rebooting the affected servers daily to clear out /var/vm in the hope that they can make it through a production day without creating too much swap. This works only sometimes, presumably depending on the load placed on the server on the given day.

Configuration:
various hardware, Mac pro & mini server. 10.8.4. 4GB RAM, 8GB RAM, 12GB RAM, doesn't seem to make any difference how much physical ram is in the machine.