aurora-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Zameer Manji <zma...@apache.org>
Subject Re: Review Request 58053: Reliably subscribe to Mesos in the HTTP Driver.
Date Fri, 31 Mar 2017 00:22:57 GMT

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/58053/
-----------------------------------------------------------

(Updated March 30, 2017, 5:22 p.m.)


Review request for Aurora and Stephan Erb.


Changes
-------

Feedback.


Bugs: AURORA-1911
    https://issues.apache.org/jira/browse/AURORA-1911


Repository: aurora


Description
-------

As noted in AURORA-1911 the `V1Mesos` driver doesn't re try `SUBSCRIBE` calls if they fail.
This means that after a leader subscribes and disconnects, it is possible for it to never
re subscribe again if the Mesos Master is unhealthy.

To fix this, I have moved the subscription into the dedicated `SchedulerExecutor` and it coninutes
to attempt to subscribe using truncated binary backoff. It only stops if we are disconnected
or if we sucessfully connect.


Diffs (updated)
-----

  src/jmh/java/org/apache/aurora/benchmark/StatusUpdateBenchmark.java 206b11458da2b0f938f0fcab5e5d3259a88ac9ee

  src/main/java/org/apache/aurora/scheduler/mesos/MesosCallbackHandler.java 5bf1e4e8c46044cb69b266cd203b5ec2f8b9ab61

  src/main/java/org/apache/aurora/scheduler/mesos/SchedulerDriverModule.java 10d4f1b515b91d85b283cb7c655275c22fb133f9

  src/main/java/org/apache/aurora/scheduler/mesos/VersionedMesosSchedulerImpl.java 67d356ab66c926a3b56860b906a453d57d6b694d

  src/test/java/org/apache/aurora/scheduler/mesos/VersionedMesosSchedulerImplTest.java 756d0d9e30a447f9fba75c1c60f2f2f3c610399b



Diff: https://reviews.apache.org/r/58053/diff/3/

Changes: https://reviews.apache.org/r/58053/diff/2-3/


Testing
-------


Thanks,

Zameer Manji


Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message