thinkharderdev opened a new pull request, #667:
URL: https://github.com/apache/arrow-ballista/pull/667
# Which issue does this PR close?
<!--
We generally require a GitHub issue to be filed for all bug fixes and
enhancements and this helps us generate change logs for our releases. You can
link an issue to this PR using the GitHub syntax. For example `Closes #123`
indicates that this PR will close issue #123.
-->
Closes #655
# Rationale for this change
<!--
Why are you proposing this change? If this is already explained clearly in
the issue then this section is not needed.
Explaining clearly why changes are proposed helps reviewers understand your
changes and offer better suggestions for fixes.
-->
See #655 issues description.
# What changes are included in this PR?
<!--
There is no need to duplicate the description in the issue here but it is
sometimes worth providing a summary of the individual changes in this PR.
-->
1. Add a new executor status `Terminating` to indicate that an executor has
started the termination process.
2. On receiving a `SIGTERM` the executor will immediately send a heartbeat
to update it's status to `Terminating`
3. After receiving a `SIGTERM` the executor should attempt to complete any
tasks currently running before terminating
4. The scheduler should consider executors in `Active` status to be live
(i.e. eligible to schedule tasks on)
5. When receiving an `executor_stopped` rpc, the scheduler will wait a
configurable amount of time before removing the executor. This can be set
through the new `--executor-termination-grace-period` cli flag and has a
default value of 30s.
6. When periodically checking for dead executors, the scheduler should also
remove any executors in `Terminating` status for longer than
`executor_termination_grace_period`.
All together this tries to ensure:
1. When an executor is terminated it has a chance to finish any in-flight
tasks to minimize disruption
2. The termination of an executor will be propagated to all schedulers in
the cluster through the heartbeat mechanism
3. The scheduler will give the terminating executor a grace period to
complete it's tasks before resetting tasks/stages and potentially recomputing
results.
# Are there any user-facing changes?
<!--
If there are user-facing changes then we may require documentation to be
updated before approving the PR.
-->
A new cli flag `--executor-termination-grace-period` for the scheduler with
a default of 30 seconds.
<!--
If there are any breaking changes to public APIs, please add the `api
change` label.
-->
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
