On Fri, 31 Mar 2006, Pierangelo Masarati wrote:
Seriously, my own workaround (do a search against the meta every five
minutes) seems to be working. I haven't specified any of those timeouts
at all, so they would be whatever the defaults are.
By default, they're off; it might be the remote server's fault if it's
doing some sort of idletimeout.
The remote server is the local server; it looks itself up via a named
pipe.
I've managed to reproduce it 100%, and it's really strange...
The FreeBSD port of OpenLDAP comes with a script
/usr/local/etc/rc.d/slapd.sh which handles starting/stopping etc. Call
this from CRON, it will fail 100% of the time, on the first search. Call
it via a remote shell, ditto. Call it from the command line, works OK.
And only on this server, which has an identical configuration to another
one.
As it would not appear to be an OpenLDAP problem after all (but perhaps a
bug is being tickled somehow) I won't pursue it any further on this list.