Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
On Tue, Jul 05, 2011 at 01:16:32PM +0200, Bart Coninckx wrote: > On 07/05/11 12:38, Pawel Warowny wrote: > >Hi > > > >Is there any way to configure cluster (pacemaker), so that some > >resource start only after drbd full synchronisation? > > > >My configuration looks like this: > >Two nodes with kvm virtualized guests. > > > >node1: > >drbd -> dlm -> some other resources -> libvirtd -> > >-> VirtualDomain1, VirtualDomain2, (...) > > > >node2 > >drbd -> dlm -> some other resources -> libvirtd -> > >-> VirtualDomain5, VirtualDomain6, (...) > > > >VirtualDomains can move between nodes in case of some node is > >unavailable. > > > >What Am I worry about: > > > >- some bad thing happened on node2 > >- node1 stonith node2, so node2 is rebooted > >- VirtualDomains from node2 migrate to node1 > >- node2 started after reboot and starting cluster resources > >- node2: drbd resync and getting primary before full sync. > >- after drbd promotion all other resources is starting on node2 > >- VirtualDomains which migrated after stonith now is going back to node2 > > > >and here is the problem, I don't want this migration while drbd is > >syncing. It leads to heavily I/O load while drbd is syncing and at the > >same time VirutalDomains are migrating. Under heavy I/O load migration > >is slow and inefficient . > > > >Even worse, while drbd is syncing and VirtualDomains are migrating > >under heavy load some VirtualDomain migration failed and because > >of that node2 stonith node1. > > > >In this situation node1 is rebooting and we have inconsistent drbd > >resource on node2 so every resource above drbd will fail. > >Node1 started after reboot and again stonith node2 because failed > >resources on it. > > > >I know that I can configure resource stickiness which will prevent > >migration but this auto migration is important feature for me. > > > >Best regards > > I don't think the resource will become primary before having fully synced. Configure DRBD resource level fencing, using e.g. the crm-fence-peer handler. -- : Lars Ellenberg : LINBIT | Your Way to High Availability : DRBD/HA support and consulting http://www.linbit.com DRBD® and LINBIT® are registered trademarks of LINBIT, Austria. __ please don't Cc me, but send to list -- I'm subscribed