Releases: criteo-forks/marathon
Releases · criteo-forks/marathon
v1.9.109-criteo25
Allow multiple health check shields per task id * Allow multiple health check shields per task id Shield is now identified by /v2/shield/<task_id>/<shield_name> * Renames after live review * Fixed tests * Better error handling in the actor
v1.9.109-criteo24
Fix a log uppon offer receival It's better to have hostname on same line, otherwise multiline logs are a real pain to parse...
v1.9.109-bes-ivnestigation-criteo1
Fix log presentation Putting back-off delay at the end makes it unlikely to see it because when there are too many scheduledInstances, log get truncated by rsyslog. Delay is more important and having an extract of few scheduledInstances is enough.
v1.9.109-oneline-criteo1
Put logs for unsatisfied offers in one line Before, log was spread accross multiple lines, and it was making investigation very difficult for incidents related to offer matching
v1.9.109-criteo23
Put logs for unsatisfied offers in one line Before, log was spread accross multiple lines, and it was making investigation very difficult for incidents related to offer matching
v1.9.109-criteo20-cni1
fixed tests
v1.9.109-criteo22
Remove old comments
v1.9.109-criteo21
Put runspec id along with logs
v1.9.109-criteo20
Let HealthCheckActor forget outdated healthcheck result We observed an issue with Mesos healthchecks. When a task failed its Mesos HC, it was killed and replaced by a new task - this was expected. What was unexpected however was that the new task was also seen as unhealthy (even if it was running correctly), with 2 healthcheck results attached to it (an outdated one reporting a failure, and a correct one reporting success). This commit fixes the issue by letting the HealthCheckActor forget outdated healthckeck result. JIRA: MESOS-4437
v1.9.109-criteo19
Let TaskReplaceActor ignore outdated health checks We observed issues when mixing Marathon and Mesos health checks. Sometimes, the Mesos health checks get duplicated, and one 'replica' keeps saying the instance is not alive, even if it is. These broken/outdated health checks can be spotted easily, since their last update timestamp is prior to the timestamp at which the task was staged at. Ignoring these outdated health checks allows the TaskReplaceActor to correctly kill the healthy instances, and perform the deployment of the new version of the application. JIRA: MESOS-4437