Avoid that an attempt to remove a snapshot that's actively used by
replication leads to a partially (or not) removed snapshot and locked
guest.
I decided to make the checks at the call sides, because passing the
log function and timeout to snapshot_delete felt awkward as they
would only be used for obtaining the lock.
Changes from v3:
* Unconditionally take the lock, to not race with replication job
creation and future-proofing.
* Only log in the case with the long timeout if we can't obtain
the lock quickly.
* Make message more general, because it might be another snapshot
removal operation holding the lock.
container:
Fabian Ebner (2):
partially fix #3424: vzdump: cleanup: wait for active replication
fix #3424: api: snapshot delete: wait for active replication
src/PVE/API2/LXC/Snapshot.pm | 11 ++++++++++-
src/PVE/VZDump/LXC.pm | 20 ++++++++++++++++++--
2 files changed, 28 insertions(+), 3 deletions(-)
qemu-server:
Fabian Ebner (1):
fix #3424: api: snapshot delete: wait for active replication
PVE/API2/Qemu.pm | 10 +++++++++-
1 file changed, 9 insertions(+), 1 deletion(-)
--
2.30.2
_______________________________________________
pve-devel mailing list
pve-devel@lists.proxmox.com
https://lists.proxmox.com/cgi-bin/mailman/listinfo/pve-devel
