[ https://issues.apache.org/jira/browse/CLOUDSTACK-8628?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14629175#comment-14629175 ]
ASF GitHub Bot commented on CLOUDSTACK-8628: -------------------------------------------- Github user bhaisaab commented on a diff in the pull request: https://github.com/apache/cloudstack/pull/580#discussion_r34755639 --- Diff: plugins/hypervisors/kvm/src/com/cloud/hypervisor/kvm/resource/wrapper/LibvirtFenceCommandWrapper.java --- @@ -48,6 +48,18 @@ public Answer execute(final FenceCommand command, final LibvirtComputingResource final KVMHAMonitor monitor = libvirtComputingResource.getMonitor(); final List<NfsStoragePool> pools = monitor.getStoragePools(); + + /** + * We can only safely fence off hosts when we use NFS + * On NFS primary storage pools hosts continuesly write + * a heartbeat. Disable Fencing Off for hosts without NFS + */ + if (pools.size() == 0) { + String logline = "No NFS storage pools found. No way to safely fence " + command.getVmName() + " on host " + command.getHostGuid(); --- End diff -- maybe add an alert for admin here, _alertMgr.sendAlert? > Ceph RBD only cluster with KVM does not fence properly > ------------------------------------------------------ > > Key: CLOUDSTACK-8628 > URL: https://issues.apache.org/jira/browse/CLOUDSTACK-8628 > Project: CloudStack > Issue Type: Bug > Security Level: Public(Anyone can view this level - this is the > default.) > Reporter: Wido den Hollander > Assignee: Wido den Hollander > Fix For: Future, 4.6.0 > > > When a NFS storage pool is present on a KVM cluster that pool is used for > writing heartbeats to. > In case of a Ceph (RBD) only cluster such pools are not available and thus > the KVM hosts do not write any heartbeats. > Should the KVM Agent crash for any reason they will be fenced of directly and > this will cause a split-brain situation. > Fencing should not happen when no NFS storage pool is present as it is > dangerous for data consistency. A Instance could be running on two hosts at > the same time. -- This message was sent by Atlassian JIRA (v6.3.4#6332)