** Description changed:

  ===== Description =====
  
  When running rbd sparsify on a volume attached to an active virtual
  machine, the task is handed off to the QEMU process because QEMU owns
  the disk lock.Once started inside QEMU, there is no way to cancel or
  interrupt the operation.
  
  Stopping or killing the rbd sparsify command line tool only closes the
  local terminal process and the operation keeps running unhindered inside
  QEMU. Additionally, internal background mechanisms ignore cancellation
  requests during remote execution. The only way to stop the process today
  is to break QEMU's disk lock, which risks causing I/O errors or freezing
  the VM.
  
  If a sparsify operation creates heavy storage traffic or slows down the
  cluster during live space cleanup, administrators have no "emergency
  brake" to safely stop it.
- 
  
  ===== Desired Behavior =====
  
  * Allow administrators to send a cancel request directly to the QEMU
  process running the sparsify task.
  
  * The cancellation process must preserve QEMU's exclusive lock on the
  volume throughout and after the abort.
  
  * Issuing a cancel command immediately halts all further block-scanning
  and discard operations inside QEMU.
  
  * The operation cancels cleanly without causing I/O errors, drops, or
  timeouts for the guest VM.
  
  * The cancellation trigger must not cause deadlocks or guest latency.
  
  * Blocks discarded before cancellation must be handled safely.
  
  * Older Ceph client versions connecting to newer clusters (or vice
  versa) should gracefully fall back if the cancellation action is
  unrecognized.
  
- 
  ===== Proposed Solution =====
  
  Consider extending the librbd ImageWatcher RPC messaging protocol to
  send a cancellation signal to the executor process (QEMU), and update
  the RemoteProgressContext to return an abort code that cleanly stops the
  rbd sparsify operation.
  
- 
  ===== Additional Info =====
  
  Feature request filed in ceph tracker:
  https://tracker.ceph.com/issues/80568
+ 
+ Target versions for this request are Tentacle (20.2.x) and later

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2167487

Title:
  Add support to cancel QEMU-executed rbd sparsify on live VMs

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/ceph/+bug/2167487/+subscriptions


-- 
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Reply via email to