Alexander Rukletsov created MESOS-6170:
------------------------------------------
Summary: Health check grace period covers failures happening after
first success.
Key: MESOS-6170
URL: https://issues.apache.org/jira/browse/MESOS-6170
Project: Mesos
Issue Type: Improvement
Affects Versions: 1.0.0
Reporter: Alexander Rukletsov
Fix For: 1.1.0
Currently, the health check library [ignores *all*
failures|https://github.com/apache/mesos/blob/b053572bc424478cafcd60d1bce078f5132c4590/src/health-check/health_checker.cpp#L192-L197]
from the task’s start (technically from the health check library
initialization) [until after the grace period
ends|https://github.com/apache/mesos/blob/b053572bc424478cafcd60d1bce078f5132c4590/include/mesos/v1/mesos.proto#L403].
This behaviour is misleading. Once the health check succeeds for the first
time, grace period rule for failures should not be applied any more.
For example, if the grace period is set to 10 minutes, the task becomes healthy
after 1 minute and fails after 2 minutes, the failure should be treated as a
normal failure with all the consequences.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
