mesos-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Vinod Kone <vinodk...@gmail.com>
Subject Re: Review Request 58194: Hardened HTTP check tests.
Date Thu, 13 Apr 2017 00:56:58 GMT


> On April 7, 2017, 12:31 a.m., Vinod Kone wrote:
> > src/tests/check_tests.cpp
> > Line 837 (original), 839 (patched)
> > <https://reviews.apache.org/r/58194/diff/1/?file=1684600#file1684600line839>
> >
> >     Does this mean the test runs for atleast 1 second? If yes, that's unfortunate.
Also, if for whatever reason (VM slowness) it takes longer than 1s this will fail?
> >     
> >     Can you check the server is ready to serve requests some other way? Maybe a
while loop that checks that the server is up or something?
> 
> Gastón Kleiman wrote:
>     Yes, this means that the tests runs for at least 1 second, however I don't see a
way of making the delay period dynamic.
>     
>     The tests could be rewritten in such a way that status updates with an empty check
status are ignored until a non-empty check status is received. Then the test should fail if
a non-empty check status is not received within 15 seconds of having started the task....
but that wouldn't be a small change, and this patch already makes the teets much more robust.
>     
>     I'd still ship this, probably mentioning in a TODO how to make the tests more robust/quick.
> 
> Alexander Rukletsov wrote:
>     I had similar thoughts when making this change. One of the main reasons we've added
checks is to tell when a server is ready to serve requests. I'd argue that making this test
run 1s is a reasonable trade-off.

I like Gaston's suggestiong of ignoring empty health checks. Is that really hard to do? I
am really averse to adding more slow tests to the already slow test suite, esp if they are
still prone to flakiness. If it takes couple more hours to write a better test, I think we
should do that. If these tests are flaky on the CI and you want do this as a quick fix to
stabilize it, I would be ok with this patch; as long as you add a clear TODO and come back
to fix it sooner than later.


- Vinod


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/58194/#review171298
-----------------------------------------------------------


On April 4, 2017, 10:24 p.m., Alexander Rukletsov wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/58194/
> -----------------------------------------------------------
> 
> (Updated April 4, 2017, 10:24 p.m.)
> 
> 
> Review request for mesos, Gastón Kleiman and Vinod Kone.
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> Introduce 1s delay to ensure the task (HTTP server) has enough time
> to start and start serving request.
> 
> 
> Diffs
> -----
> 
>   src/tests/check_tests.cpp d7fcbf940102efe0841f07e8c4937a78361bc422 
> 
> 
> Diff: https://reviews.apache.org/r/58194/diff/1/
> 
> 
> Testing
> -------
> 
> See https://reviews.apache.org/r/58196/
> 
> 
> Thanks,
> 
> Alexander Rukletsov
> 
>


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