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

Re: back-ldap retries

On Thu, 2006-02-16 at 16:26 -0500, Aaron Richton wrote:
> Sometimes, a copy of slapd running back-ldap and glue appears hung from a
> user's persepctive. Sooner or later, I get in the log:
> conn=89 op=2 ldap_back_retry: retrying URI="ldap://ldap.nbcs.rutgers.edu"; DN=""
> and then things work again. I'm guessing that there's some sort of network
> timeout. conn=89 op=1 was at 16:10:13, the retry message was 16:13:43, and
> things looked bad from a user perspective for that 3:30.
> I'm guessing that setting idle-timeout won't help because the connection
> isn't idle, and the man page says that with the timeout keyword, "only
> write operations are addressed." Is there any way to bring down the
> "I give up and am going to retry" interval?
> note, if this sounds like a bug rather than a configurable, I have gcore
> from 16:11 available. But I think I just need a more aggressive
> back-ldap...

This message means that ldap_back_retry() was invoked.  This occurs when
an existing connection appears broken while performing an operation.
The "timeout" parameter affects all operations but searches.  Searches
are affected by the server-enforced or the client requested timelimit in
a much similar manner.  The point is that in any case that
timeout/timelimit is passed to ldap_result(), so if the hanging is
related to some tcp/ip issue, there's no guarantee that
select/poll/epoll honors it.  This might be what's occurring to you.
You might force your connections to refresh often by using idle-timeout
(invalidate a connection if it's been idle for more than some ttl) or
conn-ttl (invalidate connections after some ttl; HEAD only: merging it
in 2.3 right now).


Ing. Pierangelo Masarati
Responsabile Open Solution
OpenLDAP Core Team

SysNet s.n.c.
Via Dossi, 8 - 27100 Pavia - ITALIA
Office:   +39.02.23998309          
Mobile:   +39.333.4963172
Email:    pierangelo.masarati@sys-net.it