mirror of
https://github.com/token2/snapd.git
synced 2026-03-13 11:15:47 -07:00
9e2e5e206a
* o/snapstate: when invoked for rollback, check for any changes related to snapd When the snapd.service unit fails to be activated, an OnFailure handler will execute snap-failure which in turn starts the snapd process from the previous revision of the snap with SNAPD_REVERT_TO_REV set in its environment. It may happen that the snapd unit fails at runtime without an associated change to the snapd snap, however snap-failure is not able to detect such case, and so the snapd process started in its context would continue to run. Avoid this by extending the logic within snapd to check it if has been started by snap-failure with the intention of handling a rollback, and so whether there is a change related to snapd snap in the state. When the conditions have not been met, snapd exits and snap-failure continues to restart the snapd service. Related issues: SNAPDENG-21605 Signed-off-by: Maciej Borzecki <maciej.borzecki@canonical.com> * tests/core/snapd-failover: account for improved snap-failure behavior Signed-off-by: Maciej Borzecki <maciej.borzecki@canonical.com> * overlord: add Is() to startup error Add Is() support to startupError, so that error can be introspected at runtime. Signed-off-by: Maciej Borzecki <maciej.borzecki@canonical.com> * o/snapstate: return an explicit error from StartUp() when no recovery was detected When snapd is invoked in a context of recovery but the state does not reflect this, return an explicit error indicating that further startup should be carried out. Signed-off-by: Maciej Borzecki <maciej.borzecki@canonical.com> * daemon: return an explicit error when startup in recovery context was aborted Return an explicit error when startup in failure recovery context was aborted due to lack of operations in the state which may have triggered it. Signed-off-by: Maciej Borzecki <maciej.borzecki@canonical.com> * cmd/snapd: handle unnecessary failure recovery Gracefully handle unnecessary failure recovery but exiting with 0 status, so that snap-failure may continue with cleanup and restart. Signed-off-by: Maciej Borzecki <maciej.borzecki@canonical.com> * o/snapstate: improve the check for asserting if restart was warranted Signed-off-by: Maciej Borzecki <maciej.borzecki@canonical.com> * overlord: add managers test for handling of runtime restart with failure handling Signed-off-by: Maciej Borzecki <maciej.borzecki@canonical.com> * many: tweak names and comments Signed-off-by: Maciej Borzecki <maciej.borzecki@canonical.com> * o/snapstate: tweak unit test names Signed-off-by: Maciej Borzecki <maciej.borzecki@canonical.com> * cmd/snapd: use fmt.Fprintln Signed-off-by: Maciej Borzecki <maciej.borzecki@canonical.com> * overlord/snapstate: tweak naming Signed-off-by: Maciej Borzecki <maciej.borzecki@canonical.com> --------- Signed-off-by: Maciej Borzecki <maciej.borzecki@canonical.com>