[MAGNOLIA-6061] Shutdown leniently at clustered environments Created: 04/Feb/15  Updated: 22/Nov/16  Resolved: 21/Nov/16

Status: Closed
Project: Magnolia
Component/s: None
Affects Version/s: None
Fix Version/s: None

Type: Improvement Priority: Critical
Reporter: Zdenek Skodik Assignee: Ilgun Ilgun
Resolution: Cannot Reproduce Votes: 0
Labels: support
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified

Issue Links:
relation
Template:
Acceptance criteria:
Empty
Task DoD:
[ ]* Doc/release notes changes? Comment present?
[ ]* Downstream builds green?
[ ]* Solution information and context easily available?
[ ]* Tests
[ ]* FixVersion filled and not yet released
[ ]  Architecture Decision Record (ADR)
Date of First Response:
Sprint: Basel 71
Story Points: 8

 Description   

During updates there used to be a lot of repository events being performed. An event can be observed, but its processing delayed for a while. It's possible that the primary instance being updated generate such events that are distributed over to other instances. The problem occurs if a shutdown sequence has been issued for such instance/s in the meanwhile. When deferring event processing like this we could expect that the repository might have been shutdown in the meanwhile and don't treat that as an error, or even all deferred processing could be aborted before shutting down the repository, since anyway your only option currently is to kill the instance at that point. Logs, thread dumps or more on the context can be found at the linked ticket.



 Comments   
Comment by Federico Grilli [ 05/May/15 ]

Raised priority to critical as not having this may cause content corruption

Comment by Philipp Bärfuss [ 28/May/15 ]

We should aim to fix that for the next major release as the fix most likely needs bigger changes. It would make sense to approach it in the context of author clustering.

Depending on the found solution we can consider porting back the solution to older versions.

Comment by Jan Haderka [ 11/Feb/16 ]

Proposed solution:

  • Observation should check whether target repo of the event is still open or already shutdown. When shutdown, just ignore the events because nothing can be done about them anymore.

Or if possible, find out whether there is way to notify other nodes by the node that is executing shutdown of the repo so those other nodes can shutdown observation safely.

Comment by Ilgun Ilgun [ 08/Sep/16 ]

For the record, we do something similar in new backup module

Edit:

I actually thought the solution would be similar but it's nowhere near it

Comment by Ilgun Ilgun [ 21/Nov/16 ]

Cannot reproduce with 5.5.1 and 5.3.18 versions therefore closing it.

Generated at Mon Feb 12 04:10:56 CET 2024 using Jira 9.4.2#940002-sha1:46d1a51de284217efdcb32434eab47a99af2938b.