Hi Florian,<div><br></div><div>My bad for my first email. I meant "You recommend the use of *Heartbeat* over OpenAIS / Corosync" (combined with Pacemaker).</div><div><br></div><div>The explanations are quite clear :-)</div>
<div><br></div><div>I am working on a two-node cluster with DRBD, Pacemaker. I guess I will follow your advices and go for Heartbeat as the messaging layer.</div><div><br></div><div>Thank you and I stay tuned for any updates from LINBIT regarding Heartbeat.</div>
<div><br></div><div>Cheers,<br><br></div><div>Romain</div><div><br><div class="gmail_quote">2009/11/17 Florian Haas <span dir="ltr"><<a href="mailto:florian.haas@linbit.com">florian.haas@linbit.com</a>></span><br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;">
<div class="im">On 2009-11-17 09:47, Stefan Seifert wrote:<br>
> On Tuesday, 17. November 2009, Florian Haas wrote:<br>
><br>
>> In contrast, OpenAIS/Corosync has two issues which are likely to affect<br>
>> specifically DRBD based 2-node clusters, and happen to be largely<br>
>> irrelevant for larger or non DRBD based clusters:<br>
>><br>
>> - redundant rings are suffering from breakage;<br>
>> - cluster messaging does not support unicast.<br>
><br>
> Also it would seem that OpenAIS does not support serial links between nodes<br>
> which makes it a NOGO for me. After years of being indoctrinated (for example<br>
> by you :) that having several independent communication paths is mandatory, I<br>
> just cannot go back that easily.<br>
<br>
</div>Yes, having several independent comm paths is a good idea, and as I said<br>
OpenAIS/Corosync is currently lacking in that regard. However, we<br>
definitely do advise *against* using serial links for cluster<br>
communications, and have done so ever since we started recommending<br>
using Heartbeat 2.x in CRM mode.<br>
<br>
With Heartbeat and Pacemaker, just as with Heartbeat 2.x with built-in<br>
CRM, don't use serial links. Use redundant network links. Serial links<br>
will break once your CIB reaches a certain size, and the cluster manager<br>
decides it must push a full CIB across the wire.<br>
<br>
Cheers,<br>
<font color="#888888">Florian<br>
<br>
</font><br>_______________________________________________<br>
drbd-user mailing list<br>
<a href="mailto:drbd-user@lists.linbit.com">drbd-user@lists.linbit.com</a><br>
<a href="http://lists.linbit.com/mailman/listinfo/drbd-user" target="_blank">http://lists.linbit.com/mailman/listinfo/drbd-user</a><br>
<br></blockquote></div><br></div>