You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@mesos.apache.org by Gastón Kleiman <ga...@mesosphere.io> on 2017/02/14 12:06:14 UTC
Improving health checks validation (MESOS-6916)
Hi all,
I'd like to improve the validation of the `HealthCheck` message (MESOS-6916
<https://issues.apache.org/jira/browse/MESOS-6916>).
I want to make sure that my proposed changes won't conflict with how people
are using Mesos today, so please let me know if a validation enforcing the
following extra constraints would fail with your current health checks:
1. 'delay_seconds' must be non-negative.
2. 'interval_seconds' must be non-negative.
3. 'timeout_seconds' must be non-negative.
4. 'grace_period_seconds' must be non-negative.
5. If `type` is `HTTP[S]`, then 'HealthCheck.HTTPCheckInfo.port' must be
greater than zero.
6. If `type` is `TCP`, then 'HealthCheck.TCPCheckInfo.port' must be
greater than zero.
7. 'consecutive_failures' must be non-negative.
8. Mesos treats `consecutive_failures = 0` in the same way as
`consecutive_failures = 1` (MESOS-6833
<https://issues.apache.org/jira/browse/MESOS-6833>). I propose allowing
`consecutive_failures = 0`, but changing the semantics, so that executors
don't kill the task if it fails the health checks.
Looking forward to your replies,
-Gastón