https://bugs.openldap.org/show_bug.cgi?id=9098

--- Comment #16 from maxime.besson@worteks.com maxime.besson@worteks.com --- The issues started appearing because one of the ldap-meta backends tended to become unreachable (through a site to site VPN) during the night.

In the log I just attached, you will see that a bunch of backends become unreachable at the same time, which, from what I could gather, comes from temporary routing issues on the OpenLDAP side.

One of my log traces however shows a crash (same stack trace again):

Jan 10 13:51:29 slapd.service: Main process exited, code=killed, status=6/ABRT

happening 10 seconds after a retry, instead of immediately after it

Jan 10 13:51:19 slapd[1409]: conn=62130 op=1 meta_back_retry[7]: meta_back_single_dobind=52

And no other backends had connectivity issues at the time.

In all of these cases, the local NIC was still online but the network issues were occuring somewhere else (router, VPN gateway...)

One important, related information from my original report is that we had no crashing issues before we set up backend timeouts (but of course, when a backend went down and no timeouts were in place, searches became completely unresponsive and ended up saturating all available openldap threads)