[Date Prev][Date Next] [Chronological] [Thread] [Top]

Weird problem moving from 2.0.14 to 2.0.18 on RedHat 7.1


recently we experienced a weird problem when moving from 2.0.14 to 2.0.18
on a production machine with RedHat 7.1. The problem didn't show up on
later release RedHat 7.2.  A rebuilt slapd starting from RedHat spec file
from 2.0.18 sources hangs on requests on the eth0 device, while works
correctly on the loopback device, that is:

slapd -h 'ldap://my.i.p/' 			does not work
slapd -h 'ldap://localhost/' 			works
slapd -h 'ldap://my.i.p/ ldap://localhost/'	does not work regardless
						of which interface is called

The hang is not repeatable (it happens after a random number of searches, 
sometimes even at the first request).
It hangs of course on select(). The problem does not show up with 2.0.14. 
I noticed there were a number of changes both in networking and in thread 
management between the two releases.

It looks like OpenLDAP improvements, which are so good that 2.0.18
is marked stable and apparently has no problems on other architectures,
triggered some bug in thread/networking/other on RedHat 7.1

I'll dig into it a bit further (or switch to 7.2); however, I'd like
to hear from someone else who had similar problems.