bhouse-nexthop opened a new issue, #14645: URL: https://github.com/apache/cloudstack/issues/14645
### problem A size-only `resizeVolume` of a running system VM's volume, for example a virtual router's ROOT, is accepted and resizes the volume on the storage. The running system VM is never told, so the guest keeps seeing the old size until it next starts. The API, the database and the storage all report the new size while the VM still runs on the old one. `VolumeApiServiceImpl.orchestrateResizeVolume` only looks the VM up as a user VM (`_userVmDao.findById`). For a system VM that returns null, so the resize is sent with the instance name `none` and no host. On KVM the image is grown, but `virsh blockresize` is never run for the running domain. The per-hypervisor checks in the same method (XenServer can only resize a stopped VM's disk; on KVM the VM's host must be Up and Enabled) are skipped for the same reason. Other paths already treat system VM volumes differently: - volume snapshots of system VM volumes are refused; - attach and detach accept only user VMs; - a ROOT resize through a new disk offering is refused unless the VM is Stopped. The size-only path has no such check. ### versions 4.22, and main. ### The steps to reproduce the bug Seen on KVM with Ceph RBD primary storage: 1. Take a running virtual router whose ROOT disk offering is not size-strict. 2. Call `resizeVolume id=<router ROOT volume> size=<larger size in GB>`. 3. The call succeeds. `listVolumes` and `rbd info` show the new size. 4. On the host, `virsh domblkinfo <router> vda` still shows the old size, and so does `lsblk` in the router. The new size only appears after the router is restarted. ### What to do about it? Refuse to resize the volume of a system VM that is not Stopped, with an error telling the admin to stop it first. Resizing it while it is stopped keeps working as before, and the VM sees the new size when it starts. Doing the resize live instead would mean running the system VM's resize through the VM work queue and notifying the hypervisor, as is done for user VMs, and checking that on each hypervisor. Refusing is the smaller, safer change. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
