On Feb 6, 2010, at 12:25 AM, Kirill Stasenkov wrote: > Thanks :) It's realy obvious. But i try to understand inner process in > pacemaker without reading code I'm not dev but i can. I looked through all > materials on clusterlabs(Configuration explained, Fencing and etc) but > couldn't find any clue, why sometimes on start-up, in cluster with more than > 2-3 nodes(no-quorum-policy=stop and no stonith) 4 node can suddenly shutdown > connection to crm?
It doesn't. The crm only shuts down when someone tells it to. > I think it's all about DC. When new nodes connects it have to became DC to > push cib --sync-all globaly, am i right? But it losses it's quorum and start > make many errors it can be seen in log i've attached. So current DC send > shutdown to new node to save cib integrity. No. Thats not happening. There would be _very_ noisy ERROR messages if that was the case. > By this time cib half-updated with new node. > There is no one who can affect corosync behaviour. > > Can you advise what to read about inner pacemaker mechanism(PE, TE, CIB), how > they can be tunned. And help me with advise on my question. > Thank you in advance i am appreciate any help :) > > On 05.02.2010, at 16:09, Andrew Beekhof wrote: > >> On Fri, Feb 5, 2010 at 11:35 AM, Kirill Stasenkov >> <kirill.stasen...@gmail.com> wrote: >>> Hi i'm new with ha and test some product where pacemaker with corosync >>> available. >>> Installing next node gives this output to log. Node registered in cluster >>> but corosyc shutdown unexpectedly with "shutdown by sysadmin" >>> Log in attach. Can you advise me anything? >> >> Looks like someone told corosync to shut down... > -- Andrew _______________________________________________ Pacemaker mailing list Pacemaker@oss.clusterlabs.org http://oss.clusterlabs.org/mailman/listinfo/pacemaker
