You are viewing a plain text version of this content. The canonical link for it is here.
Posted to reviews@mesos.apache.org by Alexander Rukletsov <ru...@gmail.com> on 2018/03/19 14:14:48 UTC
Re: Review Request 65581: Made all allocator tests allow for slave
backoff.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/65581/#review199414
-----------------------------------------------------------
src/tests/master_allocator_tests.cpp
Line 175 (original), 177-178 (patched)
<https://reviews.apache.org/r/65581/#comment279742>
You definitely want to add `.WillRepeatedly(Return())` here.
src/tests/master_allocator_tests.cpp
Lines 184-186 (patched)
<https://reviews.apache.org/r/65581/#comment279743>
I'm not sure you need these lines. They ensure that a second registration attempt always happens, which is very important if we have two different agents and such registration attempt triggers a registration exepectation for the second agent. So, as a general pattern it might be useful, but technically in this test I don't see any value. Am I missing anything?
- Alexander Rukletsov
On Feb. 9, 2018, 2:28 a.m., Till Toenshoff wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/65581/
> -----------------------------------------------------------
>
> (Updated Feb. 9, 2018, 2:28 a.m.)
>
>
> Review request for mesos and Benjamin Bannier.
>
>
> Bugs: MESOS-8463
> https://issues.apache.org/jira/browse/MESOS-8463
>
>
> Repository: mesos
>
>
> Description
> -------
>
> Most of the allocator tests were expecting a single invocation of
> 'AddSave' on the allocator when registering a slave. Due to the nature
> of our slave registration backoff, a single slave possibly sends
> mutiple registration requests in short succession. Any event closely
> connected to a slave registration will therefor be possibly invoked
> multiple times over the lifetime of such test.
> We now allow multiple invocations of 'AddSlave'. This patch also
> enhances the timing control by introducing a paused clock over the
> slave registration phase of these tests.
>
>
> Diffs
> -----
>
> src/tests/master_allocator_tests.cpp e4a63836ba8cae7b9cf2fce9d46a844858749182
>
>
> Diff: https://reviews.apache.org/r/65581/diff/4/
>
>
> Testing
> -------
>
> make check && ./bin/mesos-tests.sh --gtest_filter="MasterAllocatorTest*" --gtest_repeat=1000 --gtest_break_on_failure
>
>
> Thanks,
>
> Till Toenshoff
>
>
Re: Review Request 65581: Made all allocator tests allow for slave
backoff.
Posted by Till Toenshoff <to...@me.com>.
> On March 19, 2018, 2:14 p.m., Alexander Rukletsov wrote:
> >
Thanks Alex - as discussed today, coming up with an update via https://issues.apache.org/jira/browse/MESOS-8613.
- Till
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/65581/#review199414
-----------------------------------------------------------
On Feb. 9, 2018, 2:28 a.m., Till Toenshoff wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/65581/
> -----------------------------------------------------------
>
> (Updated Feb. 9, 2018, 2:28 a.m.)
>
>
> Review request for mesos and Benjamin Bannier.
>
>
> Bugs: MESOS-8463
> https://issues.apache.org/jira/browse/MESOS-8463
>
>
> Repository: mesos
>
>
> Description
> -------
>
> Most of the allocator tests were expecting a single invocation of
> 'AddSave' on the allocator when registering a slave. Due to the nature
> of our slave registration backoff, a single slave possibly sends
> mutiple registration requests in short succession. Any event closely
> connected to a slave registration will therefor be possibly invoked
> multiple times over the lifetime of such test.
> We now allow multiple invocations of 'AddSlave'. This patch also
> enhances the timing control by introducing a paused clock over the
> slave registration phase of these tests.
>
>
> Diffs
> -----
>
> src/tests/master_allocator_tests.cpp e4a63836ba8cae7b9cf2fce9d46a844858749182
>
>
> Diff: https://reviews.apache.org/r/65581/diff/4/
>
>
> Testing
> -------
>
> make check && ./bin/mesos-tests.sh --gtest_filter="MasterAllocatorTest*" --gtest_repeat=1000 --gtest_break_on_failure
>
>
> Thanks,
>
> Till Toenshoff
>
>