On Fri, 4 Apr 2008, Howard Chu wrote:
rein@basefarm.no wrote:
So far it looks to me as if the syncrepl thread has managed to return without adding the connection socket back to slap_daemon.sd_readers. But whether that is correct or not, and how it has managed to do it if so, I cannot tell.
Since you're working with HEAD ... Were there any write operations on cn=config just before the hang? (Thinking it may be connected to the SYNC_PAUSED handling at line 1303, except any subsequent request would wake up the listener anyway.)
No, the config backend isn't even configured into the server.
I have also tried without the auditlog just to make sure that wasn't involved, but it happend anyhow.
We have also added a linux server on one of the sites to see if the problem is related to solaris. So far it hasn't stopped receiving updates, but is has only run since last evening so it is far too early to conclude anything.
Rein