mesos-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jie Yu <yujie....@gmail.com>
Subject Re: Review Request 68239: Updated port mapper CNI test.
Date Mon, 06 Aug 2018 23:13:09 GMT


> On Aug. 6, 2018, 10:35 p.m., Chun-Hung Hsiao wrote:
> > src/tests/containerizer/cni_isolator_tests.cpp
> > Lines 1799 (patched)
> > <https://reviews.apache.org/r/68239/diff/2/?file=2069278#file2069278line1799>
> >
> >     What's the reason of doing a pre-test cleanup first? To prevent residual rules
from the previous run?

Yeah, that's the typical pattern we used in the test.


> On Aug. 6, 2018, 10:35 p.m., Chun-Hung Hsiao wrote:
> > src/tests/containerizer/cni_isolator_tests.cpp
> > Lines 1929-1932 (patched)
> > <https://reviews.apache.org/r/68239/diff/2/?file=2069278#file2069278line1932>
> >
> >     Why do we need to consume a new port ID for the container port?

Because container is running on the host network `__MESOS_TEST__` network does not fork a
new network namespace.


> On Aug. 6, 2018, 10:35 p.m., Chun-Hung Hsiao wrote:
> > src/tests/containerizer/cni_isolator_tests.cpp
> > Lines 1988 (patched)
> > <https://reviews.apache.org/r/68239/diff/2/?file=2069278#file2069278line1995>
> >
> >     Nit: s/`containerId`/`containerIds`/
> >     
> >     Also, it seems we don't use `containerId` in this test once we get them at all?

Removed.


> On Aug. 6, 2018, 10:35 p.m., Chun-Hung Hsiao wrote:
> > src/tests/containerizer/cni_isolator_tests.cpp
> > Lines 1957-1958 (original), 1990-1996 (patched)
> > <https://reviews.apache.org/r/68239/diff/2/?file=2069278#file2069278line1997>
> >
> >     This test waits for all `TASK_RUNNING` before any `TASK_FINISHED`. What if the
first `TASK_FINISHED` is delivered before the last `TASK_RUNNING`?
> 
> Chun-Hung Hsiao wrote:
>     I missed the fact that the tasks won't finish until the `echo foo | nc` commands
happen. Dropping this.

I added some comment about the nc server behavior.


> On Aug. 6, 2018, 10:35 p.m., Chun-Hung Hsiao wrote:
> > src/tests/containerizer/cni_isolator_tests.cpp
> > Lines 2000-2004 (patched)
> > <https://reviews.apache.org/r/68239/diff/2/?file=2069278#file2069278line2007>
> >
> >     Is it possible to just use this loop to get the container IDs and network infos?

It's logically a bit weird to combine the loop. I'll stick to the current way.


> On Aug. 6, 2018, 10:35 p.m., Chun-Hung Hsiao wrote:
> > src/tests/containerizer/cni_isolator_tests.cpp
> > Lines 2006-2012 (patched)
> > <https://reviews.apache.org/r/68239/diff/2/?file=2069278#file2069278line2013>
> >
> >     How about setting up expectations for `lostExecutor` instead?

I'll use this for now since this is how it was the case before.


> On Aug. 6, 2018, 10:35 p.m., Chun-Hung Hsiao wrote:
> > src/tests/containerizer/cni_isolator_tests.cpp
> > Lines 1984-1987 (original), 2040-2043 (patched)
> > <https://reviews.apache.org/r/68239/diff/2/?file=2069278#file2069278line2047>
> >
> >     If we use a matcher above, and use a`vector` instead of an array we can simply
do `AWAIT_READY(collect(statusesFinished))` here. But I'm fine using loop here so not opening
an issue.

Good idea!


> On Aug. 6, 2018, 10:35 p.m., Chun-Hung Hsiao wrote:
> > src/tests/containerizer/cni_isolator_tests.cpp
> > Line 2000 (original), 2045-2047 (patched)
> > <https://reviews.apache.org/r/68239/diff/2/?file=2069278#file2069278line2063>
> >
> >     We could do `AWAIT_READY(collect(gcSchedule))` if `gcSchedule` is a vector.

Good idea!


- Jie


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/68239/#review206907
-----------------------------------------------------------


On Aug. 6, 2018, 8:31 p.m., Jie Yu wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/68239/
> -----------------------------------------------------------
> 
> (Updated Aug. 6, 2018, 8:31 p.m.)
> 
> 
> Review request for mesos, Chun-Hung Hsiao and Greg Mann.
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> This patch updated the port mapper CNI test to launch multiple
> containers concurrently. This would allow us to catch the scenarios
> where multiple iptables commands are executed concurrently.
> 
> This test fails if the fix for MESOS-9125 is not included.
> 
> 
> Diffs
> -----
> 
>   src/common/values.hpp 39487b955fc1a3c963f69de66ba0da869dd3ab2e 
>   src/tests/containerizer/cni_isolator_tests.cpp 90d2d4103c8136d2dd883318acc135f7efca80d8

> 
> 
> Diff: https://reviews.apache.org/r/68239/diff/2/
> 
> 
> Testing
> -------
> 
> sudo make check
> 
> 
> Thanks,
> 
> Jie Yu
> 
>


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