[Pacemaker] Question about cluster start-up in a 2 node cluster with a node offline.

2012-02-08 Thread James FLatten
Maybe someone can help me understand this behavior. I have a two node cluster built using the "Cluster from Scratch" document. I am testing various behaviors, one of which is to bring up the cluster services on one node while it's partner node is offline. What I am seeing is this: I start u

Re: [Pacemaker] Question about cluster start-up in a 2 node cluster with a node offline.

2012-02-10 Thread James FLatten
On 02/10/2012 04:23 PM, emmanuel segura wrote: primitive ipmi-fencing stonith:fence_ipmilan \ params pcmk_host_list="nodea nodeb" ipaddr="xxx.xxx.xxx.xxx" login="xxx" passwd="xxx" lanplus="1" timeout="4" auth="md5" \ op monitor interval="60s" This is almost exactly straight fr

Re: [Pacemaker] Question about cluster start-up in a 2 node cluster with a node offline.

2012-02-13 Thread James FLatten
On 02/12/2012 04:55 PM, Andreas Kurz wrote: just do be sure ... you also have an ipmi fencing device capable of fencing more than one node? ... like the blade center mentioned in "cluster from scratch"? Andreas- Maybe this is the problem. I am directly connecting to the iLO boards on-board ea

Re: [Pacemaker] Question about cluster start-up in a 2 node cluster with a node offline.

2012-02-13 Thread James FLatten
Andreas- As far as I can tell from the logs I just posted, after the nodea is fenced, nodeb calls 'start' on both the drbd resources. A promote is never called. Is there a way to get the drbd resource to default to promote as the call instead of start? -Davin _

Re: [Pacemaker] Question about cluster start-up in a 2 node cluster with a node offline.

2012-02-13 Thread James FLatten
On 02/13/2012 12:01 PM, emmanuel segura wrote: crm configure property startup-fencing=false Emmanuel- Is this not dangerous? I want the node to fence the other node before starting services, if it is not clean. -Davin ___ Pacemaker mailing list:

[Pacemaker] Configuration help.

2012-02-15 Thread James FLatten
I hope this mess makes sense! My current setup looks like this: nodea nodeb +---+ +---+ | eth0 |-| eth0 | | eht1 |-| eth1 | | eth2 |--+ +---| eth2 | | | | | | | | | +--| ilo

Re: [Pacemaker] Configuration help.

2012-02-15 Thread James FLatten
Never mind. I re-did my cluster fail-over tests and I did not see this behavior. It is possible that I caught the cluster during a DRBD re-sync. Thank you, -Davin On 02/15/2012 11:35 AM, James FLatten wrote: I hope this mess makes sense! My current setup looks like this: nodea

Re: [Pacemaker] Configuration help.

2012-02-16 Thread James FLatten
On 02/15/2012 11:35 AM, James FLatten wrote: I have crossed connected the HP iLO3 interfaces and setup stonith on each node. I have figured out the issue and I am posting for anyone in the future that runs into this. It turns out that the HP iLO3 will stay on-line for around 3 seconds or