v1.9.109-criteo18

Log when instances from old RunSpec are staged after TaskReplaceActor…

… start

JIRA: MESOS-4412

v1.9.109-criteo17

Ensure TaskReplaceActor sees all updates before taking decisions

JIRA: MESOS-4412

v1.9.109-criteo16

Avoid killing instances upon scale down when it is unnecessary (#39)

Example of issue that is fixed with this patch:
- context: an app has 100 green instances and 50 unscheduled ones
- trigger: the app is scaled down to 100 instances
- issue: Marathon deletes 50 green instances instead of the unscheduled ones

JIRA: MESOS-4489

Co-authored-by: Flavien Quesnel <[email protected]>

v1.9.109-criteo15

Prevent Marathon from starting tasks with old conf after force deploy…

… (#38)

We observed that, when a user deploys a version of an app which is buggy
and does not start correctly, and then forces the deployment of a new
version, Marathon continues to start tasks from the old buggy version.

This commit fixes this issue.

JIRA: MESOS-4381

Co-authored-by: Flavien Quesnel <[email protected]>

v1.9.109-criteo14

Removed one more multiline log

They take space for no good reasons

Change-Id: Ifda595b1beb10eb0274e5d2526c568ab75458459

v1.9.109-criteo13

Health check shield API

Implementation of an API which temporarly prevents the task from being killed when it's unhealthy.
This is required to support troubleshooting scenarios when the app is "legally" unhealthy, for example when taking a memory dump.

API:
- GET /v2/shield: show current shields
- PUT /v2/shield/<task_id>?ttl=<ttl>: add the shield
- DELETE /v2/shield/<task_id>: remove the shield

Regarding authorization:
- read access (GET) only require authentication
- write access (PUT & DELETE) requires UpdateRunSpec of specified task
If there's no application delete is allowed (to be able to cleanup shields of stale apps using API)

v1.9.109-criteo12

Finally fixed the bug where marathon were not releasing the lock

This was appearing because the KillStreamWatcher was waiting for all
tasks to die, and not only the overdue ones.

This was appearing most of the time when tasks were Unreachable because
they sometimes re-appear while marathon already scheduled new ones.

JIRA: MESOS-4345

v1.9.109-criteo11-rc3

Finally fixed the bug where marathon were not releasing the lock

This was appearing because the KillStreamWatcher was waiting for all
tasks to die, and not only the overdue ones.

This was appearing most of the time when tasks were Unreachable because
they sometimes re-appear while marathon already scheduled new ones.

JIRA: MESOS-4345

v1.9.109-shield-rc3

Auth switch and ttl validation

v1.9.109-criteo11-rc2

Improve even more logs to fix 4345

This:
* Fix the log to be printed when we change an already locked version
* Add logs when we wait for futures waiting for a decom
* Print in a better format locks that are already taken