Alexander Rukletsov created MESOS-7606:
------------------------------------------
Summary: Hierarchical allocator seems to perform redundant
activation / deactivation of newly added frameworks.
Key: MESOS-7606
URL: https://issues.apache.org/jira/browse/MESOS-7606
Project: Mesos
Issue Type: Bug
Components: allocation
Affects Versions: 1.3.0
Reporter: Alexander Rukletsov
According to the logs,
{noformat}
Jun 01 11:32:58 ip-172-31-6-200.us-west-2.compute.internal mesos-master[29716]:
I0601 11:32:58.226356 29728 hierarchical.cpp:274] Added framework
6dd898d6-7f3a-406c-8ead-24b4d55ed262-0018-driver-20170601113252-0092
Jun 01 11:32:58 ip-172-31-6-200.us-west-2.compute.internal mesos-master[29716]:
I0601 11:32:58.226405 29728 hierarchical.cpp:379] Deactivated framework
6dd898d6-7f3a-406c-8ead-24b4d55ed262-0018-driver-20170601113252-0092
Jun 01 11:32:58 ip-172-31-6-200.us-west-2.compute.internal mesos-master[29716]:
I0601 11:32:58.228570 29728 hierarchical.cpp:343] Activated framework
6dd898d6-7f3a-406c-8ead-24b4d55ed262-0018-driver-20170601113252-0092
{noformat}
the built-in allocator ensures that upon addition, a framework is deactivated
first and then activates it again. This seems to be redundant: if a sorter
client should always start deactivated, we should not call deactivate on it but
rather add it in a way that it is deactivated. This will naturally eliminate
the logging issue.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
