mesos-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Alexander Rukletsov <ruklet...@gmail.com>
Subject Re: Review Request 66313: Fixed an oversubscription test for agent registration backoff.
Date Tue, 27 Mar 2018 22:04:05 GMT

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/66313/#review200075
-----------------------------------------------------------


Ship it!




Ship It!

- Alexander Rukletsov


On March 27, 2018, 6:32 p.m., Benjamin Bannier wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/66313/
> -----------------------------------------------------------
> 
> (Updated March 27, 2018, 6:32 p.m.)
> 
> 
> Review request for mesos, Alexander Rukletsov and Till Toenshoff.
> 
> 
> Bugs: MESOS-8733
>     https://issues.apache.org/jira/browse/MESOS-8733
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> In the `OversubscriptionTest.ForwardUpdateSlaveMessage` test we
> observe a single `UpdateSlaveMessage` to make sure the agent has fully
> recovered. This message is sent from the resource provider-capable
> agent to communicate its (empty) set of resource providers after
> registration.
> 
> Since message was sent with a running clock, it is possible that the
> agent encounters a timeout of its registration backoff timers. The
> agent would then register agent, triggering another similar message
> which is not expected in the test.
> 
> This patch adjusts the test to always run with paused clock
> eliminating this particular scenario.
> 
> 
> Diffs
> -----
> 
>   src/tests/oversubscription_tests.cpp 47c51e3d035eb5143d00efb466675eb02236b52e 
> 
> 
> Diff: https://reviews.apache.org/r/66313/diff/1/
> 
> 
> Testing
> -------
> 
> `make check`
> 
> Before this patch this test would fail after a handful of iterations; with this patch
I was able to execute this test >110,000 times without issues.
> 
> 
> Thanks,
> 
> Benjamin Bannier
> 
>


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