Tiger Server Directory Service going down periodically

[Hanover Fisk]

I’ve recently upgraded to 10.4 server and now the Directory Service goes down every four days or so with about 500 log entries like this:

Jul 11 05:01:39 xserve DirectoryService[48]: socket(PF_ROUTE) failed: Too many open files

Web service runs, but anything requiring authentication – mail, login, ssh – is dead. The server has to be physically powered down then a cold boot. At first, it seemed to be in conjunction with a VPN login. I’ve since killed VPN in the gui and hostconfig. Now it seems to occur in conjunction with cyrus db checkpointing, though that happens routinely with no issues. PF_ROUTE is the private VPN route, but all VPN services should be dead. I have off-campus users, so I don’t want to go through re-promoting OD as users will lose their passwords. Restoring from an archive may be futile if OD had a problem to begin with. All was working perfect with 10.3 – no problems at all. DNS is properly configured but I have it prepared to come from another machine if need be. If I can get Directory Services stable, I can live without VPN services from that machine. Wish I would have waited a bit further to do the install. I thought 3 weeks would be enough time to hear of any problems with the upgrade on the discussion boards, but, apparently, thats what everyone else was thinking. Glad (in a relative sense) I’m not the only one that has had issues. Any thoughts would be appreciated.

[Anonymous]

You should try to enter following command with root privileges:

lsof

this lists the open files. you could also write a daily cronjob for checking the number of open files for a certain process or user and if this number exceeds the limit than you just send out a

shutdown -r now

regards,
helge

[Author]

Posts