Releases: criteo-forks/marathon
Releases · criteo-forks/marathon
v1.9.109-criteo6
Fix release name and tag The version extracted from the artifact misses the "v" prefix.
1.9.109-criteo5
Merge branch '1.9.109-versionning' into criteo/1.9.109
1.9.109-criteo4-3-g983ea1643
First version of TaskReplaceActor patch Summary: This patch aims to fix bad behavior for marathon when there was a leader re-election during long deployments. It used to kill too many tasks, not respecting minimumHealthCapacity. Instead of taking decisions based on events, the new Actor fetches whole state of application tasks to take decisions. Things todo left: * Take into account readiness checks * Clean SCALA warnings * Make timer tick configurable
v1.9.109-rc3
Squash of replaceactortesting branch without WorkQueue disabled * Simplify isHealthy() and remove unused methods * Fix restart strategy computation at the end of deployment * Possibly fix also other corner cases * Fix TaskReplaceActor fetching instances state * Add tools to make tests and fix 3 tests in the last testsuite * Fix some more tests, but break 1 restart strategy test * RestartStrategy needs to be reworked and seems fragile for now * Fix 2 tests in ReplaceActorTest suite * One of the test is slightly different; we don't have to wait for one instance to be dead when in overcapacity. * Improve readability and fix (?) first readiness check * Fix RestartStrategy * Attempt #1 to fix CI * Make some cleanup of unused methods * Attempt to fix CI #2: Fix fourth test * we can schedule 2 instances as there is no maximum over capacity. * Fix behavior of restartStrategy: wait for kill before start a new instance * Rename ignition to restart Strategy (more adapted to new behavior)
v1.9.109-rc2
Squash of replaceactortesting branch without WorkQueue disabled * Simplify isHealthy() and remove unused methods * Fix restart strategy computation at the end of deployment * Possibly fix also other corner cases * Fix TaskReplaceActor fetching instances state * Add tools to make tests and fix 3 tests in the last testsuite * Fix some more tests, but break 1 restart strategy test * RestartStrategy needs to be reworked and seems fragile for now * Fix 2 tests in ReplaceActorTest suite * One of the test is slightly different; we don't have to wait for one instance to be dead when in overcapacity. * Improve readability and fix (?) first readiness check * Fix RestartStrategy * Attempt #1 to fix CI * Make some cleanup of unused methods * Attempt to fix CI #2: Fix fourth test * we can schedule 2 instances as there is no maximum over capacity. * Fix behavior of restartStrategy: wait for kill before start a new instance * Rename ignition to restart Strategy (more adapted to new behavior)
1.9.109-criteo4
Fix release version identification (#29)
taskreplaceactor-bugfix-0.6
Squash of replaceactortesting branch without WorkQueue disabled * Simplify isHealthy() and remove unused methods * Fix restart strategy computation at the end of deployment * Possibly fix also other corner cases * Fix TaskReplaceActor fetching instances state * Add tools to make tests and fix 3 tests in the last testsuite * Fix some more tests, but break 1 restart strategy test * RestartStrategy needs to be reworked and seems fragile for now * Fix 2 tests in ReplaceActorTest suite * One of the test is slightly different; we don't have to wait for one instance to be dead when in overcapacity. * Improve readability and fix (?) first readiness check * Fix RestartStrategy * Attempt #1 to fix CI * Make some cleanup of unused methods * Attempt to fix CI #2: Fix fourth test * we can schedule 2 instances as there is no maximum over capacity. * Fix behavior of restartStrategy: wait for kill before start a new instance * Rename ignition to restart Strategy (more adapted to new behavior)
v1.9.109-rc1
TMP disable TaskReplaceActorTest for package generation
v1.9.126-c10a5a84d
Disable TaskReplaceActorTest for now
test-instance-id
Export MARATHON_REUSABLE_ID to both env and labels for using on client side