ejabberd - Comments for "LDAP backups configuration" https://www.ejabberd.im/node/3684 en My 2 cent: this connection https://www.ejabberd.im/node/3684#comment-54924 <p>My 2 cent: this connection client -&gt; LDAP server is with any kind of client, not only with ejd. We had to deal with 3 LDAP servers beeing used in round robin style and we finally decided to go the load balancer way. This simplified many more things, not only the jabber login.</p> Fri, 20 Nov 2009 16:20:09 +0000 ckone comment 54924 at https://www.ejabberd.im Re: LDAP backups configuration https://www.ejabberd.im/node/3684#comment-54723 <p><cite>What do you mean? There is just a simple round-robin.</cite></p> <p>In fact it seems to work but there seems to be an association between client =&gt; LDAP server. So when the LDAP server associated with a client fails, that client can't connect anymore.</p> <p>Until a fix is available I think I will stay with just one LDAP server. The behavior seems too much unpredictable for now.</p> <p>Thanks for your answers.</p> Thu, 08 Oct 2009 14:18:13 +0000 jsh comment 54723 at https://www.ejabberd.im Re: Isolate servers from client https://www.ejabberd.im/node/3684#comment-54722 <p>Humm, I think we don't have the same conception of "simple" concept ;)</p> <p>For me the simplest solution is that ejabberd behaves like we expect it to do. Use a pool of LDAP servers and mark server that are unreachable as invalid and poll them regularly to notice when they become reachable again. I'm sure that future releases of ejabberd will address that problem better. There is already a bug open concerning ldap_backups directive.</p> <p>Nevertheless your solution is very good but I can't aford it ;)</p> Thu, 08 Oct 2009 14:14:56 +0000 jsh comment 54722 at https://www.ejabberd.im Re: LDAP backups configuration https://www.ejabberd.im/node/3684#comment-54714 <p><cite>Huum, this is really an annoying and weird behavior.</cite></p> <p>That is a misfeature ;)</p> <p><cite>I need to validate the fact that only half of the requests fail when one LDAP server is down.</cite></p> <p>What do you mean? There is just a simple round-robin.</p> Thu, 08 Oct 2009 03:41:56 +0000 zinid comment 54714 at https://www.ejabberd.im Isolate servers from client https://www.ejabberd.im/node/3684#comment-54713 <p>The simplest solution is to place a load-balancer between client and servers and let the load balancer manager failover - isolation of the layers of your system. Alternatively, Directory Proxy Server can perform this same duty, and is LDAP sensitive.</p> Wed, 07 Oct 2009 23:31:12 +0000 ff1959 comment 54713 at https://www.ejabberd.im Re: LDAP backups configuration https://www.ejabberd.im/node/3684#comment-54712 <p>Huum, this is really an annoying and weird behavior.<br /> I have to choose between all requests fail (one server config) or half requests fail (two servers config).<br /> I need to validate the fact that only half of the requests fail when one LDAP server is down.</p> Tue, 06 Oct 2009 15:53:34 +0000 jsh comment 54712 at https://www.ejabberd.im Re: LDAP backups configuration https://www.ejabberd.im/node/3684#comment-54711 <p>I don't think there is a simple solution of your problem because in the case of {ldap_servers, ["server1", "server2"]} round-robin is used, hence if one of the servers fails then 50% of requests fails. In the case of ldap_backups there is no switch back to a normal state (from backup server to main server).</p> Tue, 06 Oct 2009 07:47:05 +0000 zinid comment 54711 at https://www.ejabberd.im