----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/13223/#review24589 -----------------------------------------------------------
plugins/hypervisors/kvm/src/com/cloud/hypervisor/kvm/storage/LibvirtStorageAdaptor.java <https://reviews.apache.org/r/13223/#comment48610> How about use the atomic semantics provided by createNewFile to coordinate between processes? For example: lock(int wait) { File lockFile = new File(lock-file-path); while(wait) { boolean havelock = false; if (lockFile.createNewFile()) { haveLock = true; break; } Time.sleep(1000); wait--; } return haveLock; } unlock() { File lockFile = new File(path-of-lock-file); lockFile.delete(); } It's simpler version of lockfile: https://github.com/spearce/jgit/blob/master/org.eclipse.jgit/src/org/eclipse/jgit/storage/file/LockFile.java - edison su On Aug. 2, 2013, 1:03 p.m., Wei Zhou wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/13223/ > ----------------------------------------------------------- > > (Updated Aug. 2, 2013, 1:03 p.m.) > > > Review request for cloudstack, edison su and Wido den Hollander. > > > Bugs: CLOUDSTACK-2729 > > > Repository: cloudstack-git > > > Description > ------- > > The storage pool issue (CLOUDSTACK-2729) is because of a bug in libvirt > (https://bugzilla.redhat.com/show_bug.cgi?id=977706) > We need to prevent deleting a volume when refreshing the pool. > This patch use a simple file lock to implement it. > > PS: I have tested another file lock similar to Read/Write file lock, but it > was very unstable. > > > Diffs > ----- > > > plugins/hypervisors/kvm/src/com/cloud/hypervisor/kvm/storage/LibvirtStorageAdaptor.java > a9baa52 > > Diff: https://reviews.apache.org/r/13223/diff/ > > > Testing > ------- > > Applied on 4.0.2 and 4.0.1 > > Testing On 4.0.1 From 20,June > 3 nodes, create a VM on each node every 15 minutes. Destroy the VMs 5 minutes > later. expunge.inteval = 600 (10 minutes), expunge.worker = 2 > > > Testing On 4.0.2 From 01,July > 2 nodes, create two VMs on each node every 5 minutes. Destroy the VMs 4 > minutes later. expunge.inteval = 600 (10 minutes), expunge.worker = 2 > > > Thanks, > > Wei Zhou > >
