Re: [AMBER-Developers] AMBER Cruise Control Server

From: Ross Walker <ross.rosswalker.co.uk>
Date: Sat, 01 Dec 2012 10:41:57 -0800

>On Fri, Nov 30, 2012 at 02:04:26PM -0800, Robin Betz wrote:
>> Regarding infinite hangs on the tests, the Cruisecontrol test task has
>>been
>> set up with a time limit of 6 hours in which to complete all the tests,
>> after which it will be killed and will show up in a failure state. Given
>> that it seems to take about 2 hours at the most, I think this gives
>>plenty
>> of time for the tests to complete without having to worry about infinite
>> hangs.
>
>But the problem with an overall time limit is that the hung test
>eats up all the remaining time and the following tests are not run.
>This has a been a nuisance for several releases where for example
>testing via a batch queue does not run all the tests because of hangs.
>
>For those using mpirun (not me at osc) see -maxtime min and -cpu time.
>On the other hand a shell wrapper of mpibla for use via DO_PARALLEL
>"Should be pretty simple for some Bash whizz to figure out."

But isn't that a good thing? If someone submits something to the tree that
leads to one of the parallel tests hanging we want the commit marked as a
failure and it to remain that way until the problem is rectified. If it
just skips a hung test and carries on running the rest of the tests then
it is not immediately obvious that somebody broke something.

Although granted the way the test output is reported right is not optimal.
Only compiler failures lead to units being listed as failed right now. We
probably want to consider adding counter for example to each of the
compile option blocks showing how many tests passed and failed or some
kind of history.

I'd ultimately like to also add in some basic benchmarks and a plot that
tracks performance over time. This may need some additional hardware
though.

All the best
Ross

/\
\/
|\oss Walker

---------------------------------------------------------
| Assistant Research Professor |
| San Diego Supercomputer Center |
| Adjunct Assistant Professor |
| Dept. of Chemistry and Biochemistry |
| University of California San Diego |
| NVIDIA Fellow |
| http://www.rosswalker.co.uk | http://www.wmd-lab.org |
| Tel: +1 858 822 0854 | EMail:- ross.rosswalker.co.uk |
---------------------------------------------------------

Note: Electronic Mail is not secure, has no guarantee of delivery, may not
be read every day, and should not be used for urgent or sensitive issues.






_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Sat Dec 01 2012 - 11:00:02 PST
Custom Search