Skip to content
This repository has been archived by the owner on Oct 23, 2024. It is now read-only.

Make the HistoryActor instance-aware #4792

Closed
meichstedt opened this issue Dec 7, 2016 · 0 comments
Closed

Make the HistoryActor instance-aware #4792

meichstedt opened this issue Dec 7, 2016 · 0 comments
Assignees

Comments

@meichstedt
Copy link
Contributor

meichstedt commented Dec 7, 2016

The HistoryActor was left in an intermediary state and handles both MesosStatusUpdateEvents and InstanceChanged events. It shouldn't handle MesosStatusUpdateEvent anymore.

By handling both it's storing each task failure for a normal app twice btw. This doesn't necessarily break things since only the last failure is kept in store, but

@meichstedt meichstedt added this to the Marathon 1.4 milestone Dec 7, 2016
@meichstedt meichstedt self-assigned this Dec 7, 2016
meichstedt added a commit that referenced this issue Jan 12, 2017
Summary:
With introducing the instance abstraction, the HistoryActor started to handle both InstanceChanged events and MesosStatusUpdateEvents for storing TaskFailures. This leads to failures being stored twice if both events are translated to a TaskFailure (the second write overwrites the first because we only keep the latest TaskFailure). Mid-term, handling the MesosStatusUpdateEvent should be removed in favor of InstanceChanged, but the current implementation re provided data is not sufficient to do so and requires more changes. This patch prevents the actor from storing the failure twice by going back to handling MesosStatusUpdateEvent onl< for the time being.

Related to #4792

Test Plan: sbt test

Reviewers: timcharper, aquamatthias, jasongilanfarr

Reviewed By: aquamatthias, jasongilanfarr

Subscribers: jenkins, marathon-team

Differential Revision: https://phabricator.mesosphere.com/D411
@meichstedt meichstedt modified the milestones: Next, Marathon 1.4 Jan 23, 2017
aquamatthias pushed a commit that referenced this issue Jan 25, 2017
Summary:
With introducing the instance abstraction, the HistoryActor started to handle both InstanceChanged events and MesosStatusUpdateEvents for storing TaskFailures. This leads to failures being stored twice if both events are translated to a TaskFailure (the second write overwrites the first because we only keep the latest TaskFailure). Mid-term, handling the MesosStatusUpdateEvent should be removed in favor of InstanceChanged, but the current implementation re provided data is not sufficient to do so and requires more changes. This patch prevents the actor from storing the failure twice by going back to handling MesosStatusUpdateEvent onl< for the time being.

Related to #4792

Test Plan: sbt test

Reviewers: timcharper, aquamatthias, jasongilanfarr

Reviewed By: aquamatthias, jasongilanfarr

Subscribers: jenkins, marathon-team

Differential Revision: https://phabricator.mesosphere.com/D411

(cherry picked from commit bd1c69a)
@marcomonaco marcomonaco modified the milestones: Backlog, Next Feb 6, 2017
@d2iq-archive d2iq-archive locked and limited conversation to collaborators Mar 27, 2017
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants