[Pacemaker] How to prevent a node that joins the cluster after reboot from starting the resources.

2011-08-16 Thread ihjaz Mohamed
Hi All, Am getting an unmanaged error as shown below when one of the node is rebooted and comes back to join the cluster. Online: [ aceblr101.com aceblr107.com ]  Resource Group: HAService      FloatingIP (ocf::heartbeat:IPaddr2):       Started aceblr107.com (unmanaged) FAILED      acesta

Re: [Pacemaker] After pacemaker is stopped on DC, it gets fenced because last 'stop' operation is lost

2011-08-16 Thread Andrew Beekhof
On Tue, Aug 16, 2011 at 7:21 PM, Vladislav Bogdanov wrote: > Hi, > > I recently discovered that when I stop pacemaker on DC node (for > upgrade), that node is shortly fenced by a new DC. > > Fencing is caused by not-stopped (from the PoV of new DC) clone instance > (dlm in my case) which was actua

Re: [Pacemaker] Pacemaker in RHEL6.

2011-08-16 Thread Andrew Beekhof
Thanks, I've added the details to the wiki page. On Tue, Aug 16, 2011 at 11:30 PM, Alex Forster wrote: > Andrew Beekhof writes: > >> I've written this up for the wiki: >>    http://clusterlabs.org/wiki/RHEL > > For reference, I ran into a similar problem, except that I have the further > constra

Re: [Pacemaker] When I started a resource after a stop in the environment that set placement-strategy, there is a problem.

2011-08-16 Thread Yuusuke IIDA
Hi, Andrew I confirmed that the problem that I reported was settled. Many thanks! Yuusuke (2011/08/15 15:10), Andrew Beekhof wrote: Here you go: http://hg.clusterlabs.org/pacemaker/devel/rev/b88e6d03d52f On Mon, Aug 15, 2011 at 4:01 PM, Yuusuke IIDA wrote: Hi, Andrew Thank you for a r

Re: [Pacemaker] Split-brain after

2011-08-16 Thread Alex Forster
In case anybody comes across this on Google, the solution for me was- In /etc/corosync/corosync.conf, enable the "ring recovery protocol"- > totem { > > ... > > rrp_mode: active > > ... > } Additionally, my IPaddr2's (with their near-instant start/stop times) were reaching an INFINITY failcount

Re: [Pacemaker] Pacemaker in RHEL6.

2011-08-16 Thread Alex Forster
Andrew Beekhof writes: > I've written this up for the wiki: >http://clusterlabs.org/wiki/RHEL For reference, I ran into a similar problem, except that I have the further constraint that I can't use outside repositories. As mentioned in this thread, the RHEL 6.x install media does come with

Re: [Pacemaker] wchih user's path is called from getenv("PATH")/external.c

2011-08-16 Thread Junko IKEDA
Hi, > The plugins path is fixed. You cannot have more than one. I'm not > sure, but it could be that it's possible to have subdirectories, > sth like external/mystuff/ipmi. sorry, my explanation was not clear. I want to use the specify "ipmitool" command, like /tmp/ipmitool not /usr/bin/ipmitool.

Re: [Pacemaker] wchih user's path is called from getenv("PATH")/external.c

2011-08-16 Thread Dejan Muhamedagic
Hi, On Mon, Aug 15, 2011 at 06:22:23PM +0900, Junko IKEDA wrote: > Hi, > > I am trying to setup external/ipmi configuration, > and binary path for ipmi is "/tmp/impi", for example. The plugins path is fixed. You cannot have more than one. I'm not sure, but it could be that it's possible to have

[Pacemaker] After pacemaker is stopped on DC, it gets fenced because last 'stop' operation is lost

2011-08-16 Thread Vladislav Bogdanov
Hi, I recently discovered that when I stop pacemaker on DC node (for upgrade), that node is shortly fenced by a new DC. Fencing is caused by not-stopped (from the PoV of new DC) clone instance (dlm in my case) which was actually stopped: Aug 16 08:56:29 v03-a pengine: [2083]: WARN: custom_action