executor recovers running tasks on failover/upgrade #166

jdef · 2015-03-10T13:48:37Z

consider the executor change: set the annotation/config-source to "task-{taskid}"; this allows executor to recover pod/task bindings on restart (perhaps for upgrade purposes?)
- this is only relevant if/when an executor process terminates; right now the executor clears all k8sm docker containers on startup. it might be nice to avoid this - but mesos has probably already sent TASK_LOST so it might be confusing to the scheduler to get a TASK_RUNNING after that
- adam says there is a JIRA talking about ExecutorLost, which seems related
  - https://issues.apache.org/jira/browse/MESOS-313
other related Mesos JIRA's

jdef · 2016-02-10T05:33:53Z

somewhat related to #324

jdef added the class/enhancement label Mar 10, 2015

jdef mentioned this issue Mar 11, 2015

kubelet: revamp the pod/container naming scheme kubernetes/kubernetes#5103

Merged

jdef added integration/mesos ha area/executor labels Mar 27, 2015

jdef added this to the EPIC: Make K8s/DCOS Great milestone Aug 12, 2015

jdef added the epic/make k8s on dcos great label Oct 23, 2015

jdef removed this from the EPIC: Make K8s/DCOS Great milestone Oct 23, 2015

jdef added pqm/stability epic/K8S-163 node maintenance and removed epic/make k8s on dcos great labels Nov 16, 2015

jdef mentioned this issue Nov 16, 2015

EPIC: create mechanism to update slaves with old k8sm executor #324

Open

11 tasks

jdef removed the ha label Nov 16, 2015

Provide feedback