Re: [ClusterLabs] Important note for anyone using guest nodes and upgrading to 1.1.18
On Mon, 2017-11-20 at 18:35 +0100, Ferenc Wágner wrote: > Ken Gaillot writes: > > > This will also be of interest to distribution packagers ... > > Hi Ken, > > Do you mean that this warrants a prominent package changelog entry? > Or what else could packagers do about this? Mainly publicizing it via changelog/release notes/whatever is available. I don't think reverting it is a good idea, but that's an option for a packager as well, if preserving behavior is more important than fixing issues. -- Ken Gaillot ___ Users mailing list: Users@clusterlabs.org http://lists.clusterlabs.org/mailman/listinfo/users Project Home: http://www.clusterlabs.org Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf Bugs: http://bugs.clusterlabs.org
Re: [ClusterLabs] Important note for anyone using guest nodes and upgrading to 1.1.18
Ken Gaillot writes: > This will also be of interest to distribution packagers ... Hi Ken, Do you mean that this warrants a prominent package changelog entry? Or what else could packagers do about this? -- Thanks, Feri ___ Users mailing list: Users@clusterlabs.org http://lists.clusterlabs.org/mailman/listinfo/users Project Home: http://www.clusterlabs.org Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf Bugs: http://bugs.clusterlabs.org
[ClusterLabs] Important note for anyone using guest nodes and upgrading to 1.1.18
This will also be of interest to distribution packagers ... There is a bug fix in 1.1.18 that may expose user cases that were relying on the old (less desirable) behavior. In 1.1.18, guest nodes (e.g. VirtualDomain resources with a remote-node parameter) will now be probed for resource status just like any other node. If someone had a use case where they used a -INFINITY location constraint to keep a resource off a guest node where it can't run (e.g. due to the software not being installed), they *should* have also used the resource-discovery=never constraint option to keep Pacemaker from probing the resource there. However, it wasn't necessary, since guest probes weren't implemented. Now that they are, they will start getting probe failures if they don't have resource-discovery=never. Probes are important for two particular cases: detecting when a resource is already running outside cluster control, and re-detecting resource status after a clean-up. So, reverting the behavior would not be a good idea; the solution really is to use resource-discovery=never when appropriate. -- Ken Gaillot ___ Users mailing list: Users@clusterlabs.org http://lists.clusterlabs.org/mailman/listinfo/users Project Home: http://www.clusterlabs.org Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf Bugs: http://bugs.clusterlabs.org
