franz1981 commented on issue #2845: ARTEMIS-2336 Use zero copy to replicate
journal/page/large message file (AGAIN)
URL: https://github.com/apache/activemq-artemis/pull/2845#issuecomment-577389284
@michaelandrepearce Please wait a bit more: I've an opened issue on Netty
related to
franz1981 commented on issue #2845: ARTEMIS-2336 Use zero copy to replicate
journal/page/large message file (AGAIN)
URL: https://github.com/apache/activemq-artemis/pull/2845#issuecomment-550951027
@clebertsuconic I believe this is getting in good shape for a review,
according to the test
franz1981 commented on issue #2845: ARTEMIS-2336 Use zero copy to replicate
journal/page/large message file (AGAIN)
URL: https://github.com/apache/activemq-artemis/pull/2845#issuecomment-549300079
@wy96f I've added an additional commit to control the backlog of unflushed
Netty writes, to
franz1981 commented on issue #2845: ARTEMIS-2336 Use zero copy to replicate
journal/page/large message file (AGAIN)
URL: https://github.com/apache/activemq-artemis/pull/2845#issuecomment-548529114
@wy96f I've tried to understand what's happening for ChunkedNioFile and
probably it needs to
franz1981 commented on issue #2845: ARTEMIS-2336 Use zero copy to replicate
journal/page/large message file (AGAIN)
URL: https://github.com/apache/activemq-artemis/pull/2845#issuecomment-547275684
Ok, now it is more clear, thanks!
So ByteBufs start to be accumulated due to FileRegions
franz1981 commented on issue #2845: ARTEMIS-2336 Use zero copy to replicate
journal/page/large message file (AGAIN)
URL: https://github.com/apache/activemq-artemis/pull/2845#issuecomment-547052727
@wy96f
One quick question: on
franz1981 commented on issue #2845: ARTEMIS-2336 Use zero copy to replicate
journal/page/large message file (AGAIN)
URL: https://github.com/apache/activemq-artemis/pull/2845#issuecomment-541495785
@wy96f sadly the rate limiter algorithm in guava is not very good fore fast
paced networks
franz1981 commented on issue #2845: ARTEMIS-2336 Use zero copy to replicate
journal/page/large message file (AGAIN)
URL: https://github.com/apache/activemq-artemis/pull/2845#issuecomment-541386793
@wy96f I'll look better into this on my back, but please take a look
yourself if you have
franz1981 commented on issue #2845: ARTEMIS-2336 Use zero copy to replicate
journal/page/large message file (AGAIN)
URL: https://github.com/apache/activemq-artemis/pull/2845#issuecomment-535862167
@wy96f probably the size estimator of Netty or/and the pending writes count
is not working
franz1981 commented on issue #2845: ARTEMIS-2336 Use zero copy to replicate
journal/page/large message file (AGAIN)
URL: https://github.com/apache/activemq-artemis/pull/2845#issuecomment-535837012
To make profiling traces comparable in number of samples you should need to
limit the
franz1981 commented on issue #2845: ARTEMIS-2336 Use zero copy to replicate
journal/page/large message file (AGAIN)
URL: https://github.com/apache/activemq-artemis/pull/2845#issuecomment-535824970
Sorry I have missed that the file region=true version was failing on timeout
as well: tbh
franz1981 commented on issue #2845: ARTEMIS-2336 Use zero copy to replicate
journal/page/large message file (AGAIN)
URL: https://github.com/apache/activemq-artemis/pull/2845#issuecomment-535803060
If both cases (with/without file region) are saturating the network, why the
latter will
franz1981 commented on issue #2845: ARTEMIS-2336 Use zero copy to replicate
journal/page/large message file (AGAIN)
URL: https://github.com/apache/activemq-artemis/pull/2845#issuecomment-535786097
@wy96f thanks to have tried! It sounds strange to me: I was thinking the
reason why was
franz1981 commented on issue #2845: ARTEMIS-2336 Use zero copy to replicate
journal/page/large message file (AGAIN)
URL: https://github.com/apache/activemq-artemis/pull/2845#issuecomment-535443555
@wy96f
Good analysis :)
Yep, so the possible solutions I see are are:
- tune
franz1981 commented on issue #2845: ARTEMIS-2336 Use zero copy to replicate
journal/page/large message file (AGAIN)
URL: https://github.com/apache/activemq-artemis/pull/2845#issuecomment-535399659
Maybe on ActiveMQChannelHandler there are others events that are not
correctly propagated
franz1981 commented on issue #2845: ARTEMIS-2336 Use zero copy to replicate
journal/page/large message file (AGAIN)
URL: https://github.com/apache/activemq-artemis/pull/2845#issuecomment-535397765
@wy96f
It's strange that the back-pressure propagation fix isn't working: have you
used
franz1981 commented on issue #2845: ARTEMIS-2336 Use zero copy to replicate
journal/page/large message file (AGAIN)
URL: https://github.com/apache/activemq-artemis/pull/2845#issuecomment-534999649
@wy96f
> . So only non file packet accounts for the pending write bytes in channel,
franz1981 commented on issue #2845: ARTEMIS-2336 Use zero copy to replicate
journal/page/large message file (AGAIN)
URL: https://github.com/apache/activemq-artemis/pull/2845#issuecomment-533885763
@clebertsuconic yes, agree and I would like to run a soak test + wait @wy96f
results as well
franz1981 commented on issue #2845: ARTEMIS-2336 Use zero copy to replicate
journal/page/large message file (AGAIN)
URL: https://github.com/apache/activemq-artemis/pull/2845#issuecomment-533881207
@wy96f I remember you've provided some numbers for ARTEMIS-2336 using one of
yours load
franz1981 commented on issue #2845: ARTEMIS-2336 Use zero copy to replicate
journal/page/large message file (AGAIN)
URL: https://github.com/apache/activemq-artemis/pull/2845#issuecomment-533874887
@ehsavoie @clebertsuconic This version is not failing anymore the wildfly
tests on
20 matches
Mail list logo
