[
https://issues.apache.org/jira/browse/MESOS-8?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13127975#comment-13127975
]
Benjamin Hindman commented on MESOS-8:
--------------------------------------
To start, I am not interested in using sqlite. I'd prefer, at least for now, to
do a "best effort" task history. This seems to be a sweet spot between
complexity and value.
In particular, I think the right approach is to add a data structure to the
master that stores a cache of the last N tasks that have reached a terminal
state (i.e., finished, failed, lost, etc). Then these tasks can be exposed
through the JSON callbacks to the webui, and the webui can be updated to show
them.
For a first pass, I think fixing the size of the cache is the right approach
... when the cache is full, the oldest item is evicted and the new one
inserted. We can consider adding something more complex later.
It should be clear that this approach will mean that after a master
failure/failover this cache of tasks will be gone. Again, provided master
failure/failover is uncommon we'll still get a lot of value from just this,
without adding the complexity of storing these tasks in something like
ZooKeeper (or doing something else like it).
> Maintain a history of executed frameworks/tasks and show it on the web UI
> -------------------------------------------------------------------------
>
> Key: MESOS-8
> URL: https://issues.apache.org/jira/browse/MESOS-8
> Project: Mesos
> Issue Type: Improvement
> Reporter: Matei Zaharia
>
> This is kind of a big task but ultimately important for cluster management.
> See https://github.com/mesos/mesos/issues#issue/39 for previous discussion on
> GitHub.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
