Re: [cf-dev] CPU issue and release 2.0.0M5

[Simon Bernard <contact@xxxxxxxxxxxxxxx>]

Just to let you know, we integrate the 2.0.0-M5 and it seems there is 
still memory leak :'(.

Most of the memory space is consume by exchange about observe request 
using block2 in ExchangeByToken map.
The strange thing is most of exchanges have a response set, so it seems 
there are "complete" but not removed from the map...
I do not understand as our tests seems to check this scenario without 
memory leak.

I will continue to investigate... but I have no clues for now...


Le 06/09/2017 à 06:53, Kraus Achim (INST/ECS4) a écrit :
> Hi Simon,
>
> >      This let me think we should release a 2.0.0M5 soon.
> >      Do you have particular point you want to address/finished for this
> > 2.0.0M5 ?
> Sounds good.
> If possible, a review and merge of PR #397 would make the new sandbox "useful".
>
> Mit freundlichen Grüßen / Best regards
>
>   Achim Kraus
>
> (INST/ECS4)
> Bosch Software Innovations GmbH | Stuttgarter Straße 130 | 71332 Waiblingen | GERMANY | www.bosch-si.com
>
> Sitz: Berlin, Registergericht: Amtsgericht Charlottenburg; HRB 148411 B
> Aufsichtsratsvorsitzender: Dr.-Ing. Thorsten Lücke; Geschäftsführung: Dr.-Ing. Rainer Kallenbach, Michael Hahn
>
>
>
> -----Original Message-----
> From: cf-dev-bounces@xxxxxxxxxxx [mailto:cf-dev-bounces@xxxxxxxxxxx] On Behalf Of Simon Bernard
> Sent: Dienstag, 5. September 2017 11:28
> To: Californium (Cf) developer discussions <cf-dev@xxxxxxxxxxx>
> Subject: [cf-dev] CPU issue and release 2.0.0M5
>
> Hi,
>
>      I talked about CPU issue in production with Californium/Scandium some weeks ago.
>
>      After a long period of time the CPU increase exponentially while the traffic is constants.
>
>      Playing with CPU profiler does not reveal "big" performance issue.
>      Analysis of Heap dump and GC logs, let me thinks that this is about memory leak.
>      The GC logs show a large CPU consumption when the heap is almost full.
>
>      The heap dump show that memory leak seems to be in InMemoryExchangeStore, more precisely ExchangeByToken map.
>      Most of Exchanges in this map are about observe request using block2.
>      The issue impact 2.0.0M3 (the one we currently using) but I think also the 2.0.0M4.
>
>      The good news : it should be fixed in master since PR 369 :
> https://github.com/eclipse/californium/pull/369
>
>       This let me think we should release a 2.0.0M5 soon.
>       Do you have particular point you want to address/finished for this
> 2.0.0M5 ?
>
> Simon
>
> _______________________________________________
> cf-dev mailing list
> cf-dev@xxxxxxxxxxx
> To change your delivery options, retrieve your password, or unsubscribe from this list, visit https://dev.eclipse.org/mailman/listinfo/cf-dev
> _______________________________________________
> cf-dev mailing list
> cf-dev@xxxxxxxxxxx
> To change your delivery options, retrieve your password, or unsubscribe from this list, visit
> https://dev.eclipse.org/mailman/listinfo/cf-dev