aurora-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Mehrdad Nurolahzade <mehr...@nurolahzade.com>
Subject Review Request 55243: AURORA-1868 Evaluate multiple preemption proposals per round
Date Fri, 06 Jan 2017 01:17:53 GMT

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/55243/
-----------------------------------------------------------

Review request for Aurora, David McLaughlin and Zameer Manji.


Bugs: AURORA-1868
    https://issues.apache.org/jira/browse/AURORA-1868


Repository: aurora


Description
-------

`TaskScheduler` makes an attempt to preempt already identified candidates through `Preemptor`
when it fails to schedule one or more tasks. However, `Preemptor` currently evaluates only
one proposal per invocation. A proposal may get vetoed at this point by scheduling filters.
If a proposal fails validation the task group might get penalized by `TaskGroups` to give
`PendingTaskProcessor` some time to find new preemption candidates; despite the fact that
another proposal may already exist in `slotCache`. This penalty might result in expiration
of existing proposals in `slotCache`, hence slowing down the overall preemption process.

This patch modifies `Preemptor` so that it evaluates all existing preemption proposals before
giving up.


Diffs
-----

  src/main/java/org/apache/aurora/scheduler/preemptor/Preemptor.java 7d2903a47dacfc35f9e547ccb6c5896efe3e013f

  src/test/java/org/apache/aurora/scheduler/preemptor/PreemptorImplTest.java 40c42b1b3aa63797da8dea61732b49155034fea2


Diff: https://reviews.apache.org/r/55243/diff/


Testing
-------

```
$ ./src/test/sh/org/apache/aurora/e2e/test_end_to_end.sh

...

*** OK (All tests passed) ***

mesos-master start/running, process 12228
+ RETCODE=0
+ restore_netrc
+ mv /home/vagrant/.netrc.bak /home/vagrant/.netrc
+ true
Connection to 127.0.0.1 closed.

real	26m20.055s
user	0m1.412s
sys	0m0.705s
```


Thanks,

Mehrdad Nurolahzade


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