[
https://issues.apache.org/jira/browse/AURORA-826?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Dominic Hamon updated AURORA-826:
---------------------------------
Description:
Add backoff to the scheduler re-registration retry timeout. The idea of retry
is to resolve dropped messages, however, the majority of retries are needed due
to a backed up master, and become duplicates. On a duplicate re-registration,
the Mesos master rescinds outstanding offers.
Backoff would reduce the amount of unnecessary rescinds.
This is necessary once the Aurora moves away from the scheduler driver and
becomes a pure framework.
was:
Add backoff to the scheduler re-registration retry timeout. The idea of retry
is to resolve dropped messages, however, the majority of retries are needed due
to a backed up master, and become duplicates. On a duplicate re-registration,
the Mesos master rescinds outstanding offers.
Backoff would reduce the amount of unnecessary rescinds.
> Add backoff to the re-regsitration retry timeout
> ------------------------------------------------
>
> Key: AURORA-826
> URL: https://issues.apache.org/jira/browse/AURORA-826
> Project: Aurora
> Issue Type: Story
> Reporter: Dominic Hamon
>
> Add backoff to the scheduler re-registration retry timeout. The idea of retry
> is to resolve dropped messages, however, the majority of retries are needed
> due to a backed up master, and become duplicates. On a duplicate
> re-registration, the Mesos master rescinds outstanding offers.
> Backoff would reduce the amount of unnecessary rescinds.
> This is necessary once the Aurora moves away from the scheduler driver and
> becomes a pure framework.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
