[sage-devel] Re: Timeouts on patchbots

[Clemens Heuberger]

Am 2017-02-13 um 19:44 schrieb Jakob Kroeker:
> It is my subjective observation that timeouts on patchbots have become 
> more
> frequent. For instance, for ticket
> https://patchbot.sagemath.org/ticket/22340/
> 
> two patchbots failed due to timeout,
> 
> 
> 
> If I recall correctly, I did hit similar issues when testing sage.
> Parallelism or any additional running processes during testing  may have
> negative impact on timings.
> (e.g memory cache is shared between several applications) .
> 
> 
> As far as I know for testing the default is to use 3 parallel threads,
> see
> 
> https://wiki.sagemath.org/buildbot
> 
> Set it to 1 and see if the problem remains, run only one patchbot.
> Please report your findings.

The problems I reported were not on "my" patchbots. "My" patchbots run on
different machines anyway.


Clemens

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To post to this group, send email to sage-devel@googlegroups.com.
Visit this group at https://groups.google.com/group/sage-devel.
For more options, visit https://groups.google.com/d/optout.

[sage-devel] Re: Timeouts on patchbots

[Volker Braun]

I think there is also still an issue where openblas internal state becomes 
wedged when interrupting; I disabled multi-threading which made it much 
less frequent but it is not 100% reliable. I'm seeing random issues in 
/abvar/ occasionally on the buildbot.




On Monday, February 13, 2017 at 12:33:14 PM UTC+1, Clemens Heuberger wrote:
>
> It is my subjective observation that timeouts on patchbots have become 
> more 
> frequent. For instance, for ticket 
> https://patchbot.sagemath.org/ticket/22340/ 
> two patchbots failed due to timeout, 
>
> quasar times out with 
>
> sage: ls = J0(46)._calculate_endomorphism_generators() ; ls ## line 213 ## 
>
> ** 
> -- 
> sage -t --long src/sage/modular/abvar/abvar_ambient_jacobian.py  # Timed 
> out 
>
> whereas pbua times out with 
>
> sage: maxima._batch('10003;',batchload=False) ## line 869 ## 
>
> ** 
> -- 
> sage -t --long --warn-long 255.7 src/sage/interfaces/maxima.py  # Timed 
> out 
>
> Do we have any idea how this can happen at all? (Naively, one would 
> believe that 
> either the test works or it does not, I do not understand this quite 
> random 
> behaviour here). 
>
> I know about the recommendation to avoid high system load; but this is 
> somewhat 
> hard to control. On "my" two patchbots, I restrict myself to one or two 
> processes although sometimes I could volunteer more CPU time, but then, 
> timeouts 
> get more likely. 
>
> Can we do anything about that? Usefullness of patchbots is somewhat 
> limited with 
> that many false positives. 
>
> Regards, 
>
> Clemens 
>

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To post to this group, send email to sage-devel@googlegroups.com.
Visit this group at https://groups.google.com/group/sage-devel.
For more options, visit https://groups.google.com/d/optout.

[sage-devel] Re: Timeouts on patchbots

[Jakob Kroeker]

>
> It is my subjective observation that timeouts on patchbots have become 
> more 
> frequent. For instance, for ticket 
> https://patchbot.sagemath.org/ticket/22340/ 
> two patchbots failed due to timeout, 
>


If I recall correctly, I did hit similar issues when testing sage.
Parallelism or any additional running processes during testing  may have 
negative impact on timings.
(e.g memory cache is shared between several applications) . 


As far as I know for testing the default is to use 3 parallel threads,
see 

https://wiki.sagemath.org/buildbot

Set it to 1 and see if the problem remains, run only one patchbot.
Please report your findings.

I know about the recommendation to avoid high system load; but this is 
> somewhat 
> hard to control. 
>

The best test environment is a separate computer for running patchbot and 
nothing else.


Jakob



Am Montag, 13. Februar 2017 12:33:14 UTC+1 schrieb Clemens Heuberger:
>
> It is my subjective observation that timeouts on patchbots have become 
> more 
> frequent. For instance, for ticket 
> https://patchbot.sagemath.org/ticket/22340/ 
> two patchbots failed due to timeout, 
>
> quasar times out with 
>
> sage: ls = J0(46)._calculate_endomorphism_generators() ; ls ## line 213 ## 
>
> ** 
> -- 
> sage -t --long src/sage/modular/abvar/abvar_ambient_jacobian.py  # Timed 
> out 
>
> whereas pbua times out with 
>
> sage: maxima._batch('10003;',batchload=False) ## line 869 ## 
>
> ** 
> -- 
> sage -t --long --warn-long 255.7 src/sage/interfaces/maxima.py  # Timed 
> out 
>
> Do we have any idea how this can happen at all? (Naively, one would 
> believe that 
> either the test works or it does not, I do not understand this quite 
> random 
> behaviour here). 
>
> I know about the recommendation to avoid high system load; but this is 
> somewhat 
> hard to control. On "my" two patchbots, I restrict myself to one or two 
> processes although sometimes I could volunteer more CPU time, but then, 
> timeouts 
> get more likely. 
>
> Can we do anything about that? Usefullness of patchbots is somewhat 
> limited with 
> that many false positives. 
>
> Regards, 
>
> Clemens 
>

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To post to this group, send email to sage-devel@googlegroups.com.
Visit this group at https://groups.google.com/group/sage-devel.
For more options, visit https://groups.google.com/d/optout.