[ https://issues.apache.org/jira/browse/MESOS-4909?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15210609#comment-15210609 ]
Alexander Rukletsov edited comment on MESOS-4909 at 3/31/16 3:15 PM: --------------------------------------------------------------------- {noformat} Commit: 7ab6a478b9cef548a2470d18bd281aee5610b62a [7ab6a47] Author: Alexander Rukletsov ruklet...@gmail.com Date: 24 Mar 2016 17:30:31 CET Committer: Alexander Rukletsov al...@apache.org Commit Date: 24 Mar 2016 18:21:03 CET Introduced KillPolicy protobuf. Describes a kill policy for a task. Currently does not express different policies (e.g. hitting HTTP endpoints), only controls how long to wait between graceful and forcible task kill. Review: https://reviews.apache.org/r/44656/ {noformat} {noformat} Commit: 1fe6221aa30f35f31378433412d8cb725009bd47 [1fe6221] Author: Alexander Rukletsov ruklet...@gmail.com Date: 24 Mar 2016 17:30:42 CET Committer: Alexander Rukletsov al...@apache.org Commit Date: 24 Mar 2016 18:21:03 CET Added validation for task's kill policy. Review: https://reviews.apache.org/r/44707/ {noformat} {noformat} Commit: d13de4c42b39037c8bd8f79122e7a9ac0d82317f [d13de4c] Author: Alexander Rukletsov ruklet...@gmail.com Date: 24 Mar 2016 17:30:52 CET Committer: Alexander Rukletsov al...@apache.org Commit Date: 24 Mar 2016 18:21:03 CET Used KillPolicy and shutdown grace period in command executor. The command executor determines how much time it allots the underlying task to clean up (effectively how long to wait for the task to comply to SIGTERM before sending SIGKILL) based on both optional task's KillPolicy and optional shutdown_grace_period field in ExecutorInfo. Manual testing was performed to ensure newly introduced protobuf fields are respected. To do that, "mesos-execute" was modified to support KillPolicy and CommandInfo.shell=false. To simulate a task that does not exit in the allotted period, a tiny app (https://github.com/rukletsov/unresponsive-process) that ignores SIGTERM was used. More details on testing in the review request. Review: https://reviews.apache.org/r/44657/ {noformat} {noformat} Commit: 327f840a32a0c8e2cb05d313b5c5f4d43a206c85 [327f840] Author: Alexander Rukletsov al...@apache.org Date: 31 Mar 2016 13:27:20 CEST Used KillPolicy and shutdown grace period in docker executor. The docker executor determines how much time it allots the underlying container to clean up (via passing the timeout to the docker daemon) based on both optional task's KillPolicy and optional shutdown_grace_period field in ExecutorInfo. Review: https://reviews.apache.org/r/44660 {noformat} was (Author: alexr): {noformat} Commit: 7ab6a478b9cef548a2470d18bd281aee5610b62a [7ab6a47] Author: Alexander Rukletsov ruklet...@gmail.com Date: 24 Mar 2016 17:30:31 CET Committer: Alexander Rukletsov al...@apache.org Commit Date: 24 Mar 2016 18:21:03 CET Introduced KillPolicy protobuf. Describes a kill policy for a task. Currently does not express different policies (e.g. hitting HTTP endpoints), only controls how long to wait between graceful and forcible task kill. Review: https://reviews.apache.org/r/44656/ {noformat} {noformat} Commit: 1fe6221aa30f35f31378433412d8cb725009bd47 [1fe6221] Author: Alexander Rukletsov ruklet...@gmail.com Date: 24 Mar 2016 17:30:42 CET Committer: Alexander Rukletsov al...@apache.org Commit Date: 24 Mar 2016 18:21:03 CET Added validation for task's kill policy. Review: https://reviews.apache.org/r/44707/ {noformat} {noformat} Commit: d13de4c42b39037c8bd8f79122e7a9ac0d82317f [d13de4c] Author: Alexander Rukletsov ruklet...@gmail.com Date: 24 Mar 2016 17:30:52 CET Committer: Alexander Rukletsov al...@apache.org Commit Date: 24 Mar 2016 18:21:03 CET Used KillPolicy and shutdown grace period in command executor. The command executor determines how much time it allots the underlying task to clean up (effectively how long to wait for the task to comply to SIGTERM before sending SIGKILL) based on both optional task's KillPolicy and optional shutdown_grace_period field in ExecutorInfo. Manual testing was performed to ensure newly introduced protobuf fields are respected. To do that, "mesos-execute" was modified to support KillPolicy and CommandInfo.shell=false. To simulate a task that does not exit in the allotted period, a tiny app (https://github.com/rukletsov/unresponsive-process) that ignores SIGTERM was used. More details on testing in the review request. Review: https://reviews.apache.org/r/44657/ {noformat} > Introduce kill policy for tasks. > -------------------------------- > > Key: MESOS-4909 > URL: https://issues.apache.org/jira/browse/MESOS-4909 > Project: Mesos > Issue Type: Improvement > Reporter: Alexander Rukletsov > Assignee: Alexander Rukletsov > Labels: mesosphere > > A task may require some time to clean up or even a special mechanism to issue > a kill request (currently it's a SIGTERM followed by SIGKILL). Introducing > kill policies per task will help address these issue. -- This message was sent by Atlassian JIRA (v6.3.4#6332)