Hi! I think the solution is switching from sequential dependencies in groups to parallel dependencies, separating in ordering and colocation In crm shell it is displayed as "order any inf: first ( then1 then2 )" and "colocation any inf: ( then11 then2 ) first".
Would this help? We had a similar problem with NFS exports... Regards, Ulrich >>> Adrian Saul <[email protected]> schrieb am 10.05.2016 um 10:06 >>> in Nachricht <[email protected]>: > Hi, > I am building up a service for doing ALUA based iSCSI targets using > pacemaker for configuration management and failover of target groups. This > is done using our own developed resource scripts. > To keep the config clean I have organised resources into groups - as I create > resources they are added to the groups, and the groups are either clone or > master groups to apply the configuration across the cluster: > > devicegroups - master/slave group for controlling the ALUA states > targets - iSCSI targets - clone group > devices - iSCSI backing devices - clone group > hostgroups - iSCSI hostgroups and LUN mappings - clone group > > There are constraints that order the devicegroup groups then targets. > Targets then hostgroups and devices then hostgroups. I have enabled > interleaving as well. > > This works reasonably well with two issues. The first is that a > configuration error in one resource can take down the entire group rather > than just the faulty resource - for example a typo in the attributes for a > device resource pointing to a non-existing device can cause it to error, > taking down the device group and by order constraints the hostgroups too. > This makes a small mistake a big one as devices go offline unnecessarily. > > The second is that it appears deleting a resource from the group (using pcs > resource delete) causes the policy engine to stop the entire group and > dependencies, remove the resource, then start them all again - which is very > disruptive to iSCSI traffic. For example if I remove a device that is no > longer in use by a hostgroup, it triggers a stop of the devices and > hostgroups groups, removes the device, then starts them all again. > > Is there a way I can avoid these two behaviours maintaining this simpler > group level config, or am I simply using them the wrong way? > > The alternative I am guessing would be to make each resource its own clone > instead, and do individual resource constraints more finely rather than > wholesale group constraints, which is possible but would require some more > careful configuration management to ensure constraints and resources are > properly associated. > > Thanks, > > Adrian > > > > > > > > > > Confidentiality: This email and any attachments are confidential and may be > subject to copyright, legal or some other professional privilege. They are > intended solely for the attention and use of the named addressee(s). They may > only be copied, distributed or disclosed with the consent of the copyright > owner. If you have received this email by mistake or by breach of the > confidentiality clause, please notify the sender immediately by return email > and delete or destroy all copies of the email. Any confidentiality, privilege > or copyright is not waived or lost because this email has been sent to you by > mistake. > > _______________________________________________ > Users mailing list: [email protected] > http://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 _______________________________________________ Users mailing list: [email protected] http://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
