add LoadSnapshotFailed in snapshot protocol, #21842 #21848
Conversation
|
Refs #21842 |
|
If the snapshot itself is corrupt (and hence prevents recovery) then one can still just delete it? |
|
LGTM |
|
yes, you can delete the snapshot if you know that you have events that cover it. It's also possible to start a PersistentActor without using snapshots: |
|
Ok, thanks for confirming. |
|
Test PASSed. |
Great, that's what I vaguely remembered - that's all we need then 👍 |
| * i.e. there was no snapshot. If snapshot could not be loaded the `Future` | ||
| * should be completed with failure. That is important because events may | ||
| * have been deleted and just replaying the events might not result in a valid | ||
| * state. |
| case Some(md) ⇒ | ||
| Try(withInputStream(md)(deserialize)) match { | ||
| case Success(s) ⇒ Some(SelectedSnapshot(md, s.data)) | ||
| case Success(s) ⇒ | ||
| Success(Some(SelectedSnapshot(md, s.data))) | ||
| case Failure(e) ⇒ | ||
| log.error(e, s"Error loading snapshot [${md}]") |
There was a problem hiding this comment.
Maybe make it: log.error(e, "Error loading snapshot [{}], remaining attempts: {}", md, remaining)?
|
Mention of this new behaviour in the docs (along with the |
|
TCK will be hard, we can't simulate deserialization failure because the choice of serialization is up to the journal implementation |
|
Ah, didn't think of that. I just realized we discussed the same thing about journal deserialization problems coverage recently. Thanks for clarifying a second time then. |
|
No problem, it's not obvious (and not optimal from testing perspective). |
|
@johanandren I have added the additional documentation, see 7ee86ae |
|
Test PASSed. |
|
Test PASSed. |
* treat snapshot load failure in same way as other recovery failures * if load of snapshot fails the persistent actor will be stopped, since we can't assume that a consistent state would be recovered just by replaying all events, since events may have been deleted * additional recovery docs * improve log message
4de7317 to
ea84b4b
Compare
we can't assume that a consistent state would be recovered just by
replaying all events, since events may have been deleted