You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@ignite.apache.org by Ilya Kasnacheev <il...@gmail.com> on 2020/02/17 12:00:08 UTC
Let's fix number of essential thread pools' threads in tests!
Hello!
I have created the following ticket:
https://issues.apache.org/jira/browse/IGNITE-12663
Our current number of threads in pools is max(8, NCPU), this means, we will
see different behavior depending on hardware.
I have adjusted GridAbstractTest so the default for most pools is 5,
excluding striped, which is 8.
5 is smallest viable number, I hope that TC suites will run slightly
faster, produce less logs, and will be easier to debug (smaller batches of
threads to scroll)
Tests pass after fixing a single test. WDYT?
Regards,
--
Ilya Kasnacheev
Re: Let's fix number of essential thread pools' threads in tests!
Posted by Ilya Kasnacheev <il...@gmail.com>.
Hello!
I have ran nightlies, it runs OK (with exception of MVCC Cache 7 which seem
to always hang on master) but there's virtually no speed difference in this
run.
Still, I think it makes sense to fix number of threads used. WDYT? I would
be glad to have a formal review.
Regards,
--
Ilya Kasnacheev
вт, 18 февр. 2020 г. в 14:53, Ilya Kasnacheev <il...@gmail.com>:
> Hello!
>
> A single runAll of this code was 5h faster than regular master, in net
> time, but this might as well be measurement artifact. I will compare Run
> All Nightly which I have already started.
>
> Regards,
> --
> Ilya Kasnacheev
>
>
> пн, 17 февр. 2020 г. в 18:04, Ivan Pavlukhin <vo...@gmail.com>:
>
>> Ilya,
>>
>> > 5 is smallest viable number, I hope that TC suites will run slightly
>> > faster, produce less logs, and will be easier to debug (smaller batches
>> of
>> > threads to scroll)
>>
>> It would be quite interesting for me to measure this positive effect.
>> Is it possible?
>>
>> From a practical standpoint I suppose it is necessary to ensure that
>> everything is ok with nightly tests as well.
>>
>> Best regards,
>> Ivan Pavlukhin
>>
>> пн, 17 февр. 2020 г. в 15:00, Ilya Kasnacheev <ilya.kasnacheev@gmail.com
>> >:
>> >
>> > Hello!
>> >
>> > I have created the following ticket:
>> >
>> > https://issues.apache.org/jira/browse/IGNITE-12663
>> >
>> > Our current number of threads in pools is max(8, NCPU), this means, we
>> will
>> > see different behavior depending on hardware.
>> >
>> > I have adjusted GridAbstractTest so the default for most pools is 5,
>> > excluding striped, which is 8.
>> >
>> > 5 is smallest viable number, I hope that TC suites will run slightly
>> > faster, produce less logs, and will be easier to debug (smaller batches
>> of
>> > threads to scroll)
>> >
>> > Tests pass after fixing a single test. WDYT?
>> >
>> > Regards,
>> > --
>> > Ilya Kasnacheev
>>
>
Re: Let's fix number of essential thread pools' threads in tests!
Posted by Ilya Kasnacheev <il...@gmail.com>.
Hello!
A single runAll of this code was 5h faster than regular master, in net
time, but this might as well be measurement artifact. I will compare Run
All Nightly which I have already started.
Regards,
--
Ilya Kasnacheev
пн, 17 февр. 2020 г. в 18:04, Ivan Pavlukhin <vo...@gmail.com>:
> Ilya,
>
> > 5 is smallest viable number, I hope that TC suites will run slightly
> > faster, produce less logs, and will be easier to debug (smaller batches
> of
> > threads to scroll)
>
> It would be quite interesting for me to measure this positive effect.
> Is it possible?
>
> From a practical standpoint I suppose it is necessary to ensure that
> everything is ok with nightly tests as well.
>
> Best regards,
> Ivan Pavlukhin
>
> пн, 17 февр. 2020 г. в 15:00, Ilya Kasnacheev <il...@gmail.com>:
> >
> > Hello!
> >
> > I have created the following ticket:
> >
> > https://issues.apache.org/jira/browse/IGNITE-12663
> >
> > Our current number of threads in pools is max(8, NCPU), this means, we
> will
> > see different behavior depending on hardware.
> >
> > I have adjusted GridAbstractTest so the default for most pools is 5,
> > excluding striped, which is 8.
> >
> > 5 is smallest viable number, I hope that TC suites will run slightly
> > faster, produce less logs, and will be easier to debug (smaller batches
> of
> > threads to scroll)
> >
> > Tests pass after fixing a single test. WDYT?
> >
> > Regards,
> > --
> > Ilya Kasnacheev
>
Re: Let's fix number of essential thread pools' threads in tests!
Posted by Ivan Pavlukhin <vo...@gmail.com>.
Ilya,
> 5 is smallest viable number, I hope that TC suites will run slightly
> faster, produce less logs, and will be easier to debug (smaller batches of
> threads to scroll)
It would be quite interesting for me to measure this positive effect.
Is it possible?
From a practical standpoint I suppose it is necessary to ensure that
everything is ok with nightly tests as well.
Best regards,
Ivan Pavlukhin
пн, 17 февр. 2020 г. в 15:00, Ilya Kasnacheev <il...@gmail.com>:
>
> Hello!
>
> I have created the following ticket:
>
> https://issues.apache.org/jira/browse/IGNITE-12663
>
> Our current number of threads in pools is max(8, NCPU), this means, we will
> see different behavior depending on hardware.
>
> I have adjusted GridAbstractTest so the default for most pools is 5,
> excluding striped, which is 8.
>
> 5 is smallest viable number, I hope that TC suites will run slightly
> faster, produce less logs, and will be easier to debug (smaller batches of
> threads to scroll)
>
> Tests pass after fixing a single test. WDYT?
>
> Regards,
> --
> Ilya Kasnacheev