RE: Checkpoints and catch-up burst (heavy back pressure)

Hi, I like the idea, will give it a try. Thanks, Arnaud De : Stephen Connolly Envoyé : mardi 5 mars 2019 13:55 À : LINZ, Arnaud Cc : zhijiang ; user Objet : Re: Checkpoints and catch-up burst (heavy back pressure) On Tue, 5 Mar 2019 at 12:48, Stephen Connolly mailto:stephen.alan.conno

Re: Checkpoints and catch-up burst (heavy back pressure)

to >> know the “load” of the app, or to determine if checkpointing takes too much >> time, so that I can do it only on purpose? >> >> >> >> Thanks, >> >> Arnaud >> >> >> >> *De :* zhijiang >> *Envoyé :* vendredi 1 mars 201

Re: Checkpoints and catch-up burst (heavy back pressure)

way, programmatically, to > know the “load” of the app, or to determine if checkpointing takes too much > time, so that I can do it only on purpose? > > > > Thanks, > > Arnaud > > > > *De :* zhijiang > *Envoyé :* vendredi 1 mars 2019 04:59 > *À :* user ; LI

Re: Checkpoints and catch-up burst (heavy back pressure)

7;origine > De : Ken Krugler > Date : ven., mars 01, 2019 7:05 PM +0100 > A : "LINZ, Arnaud" > CC : zhijiang , user > Objet : Re: Checkpoints and catch-up burst (heavy back pressure) > > Hi Arnaud, > > 1. What’s your checkpoint configuration? Wond

Re: Checkpoints and catch-up burst (heavy back pressure)

01, 2019 7:05 PM +0100 A : "LINZ, Arnaud" CC : zhijiang , user Objet : Re: Checkpoints and catch-up burst (heavy back pressure) Hi Arnaud, 1. What’s your checkpoint configuration? Wondering if you’re writing to HDFS, and thus the load you’re putting on it while catching up & ch

Re: Checkpoints and catch-up burst (heavy back pressure)

er Objet : RE: Checkpoints and catch-up burst (heavy back pressure) Hi, I think I should go into more details to explain my use case. I have one non parallel source (parallelism = 1) that list binary files in a HDFS directory. DataSet emitted by the source is a data set of file names, not file content.

RE: Checkpoints and catch-up burst (heavy back pressure)

Arnaud Objet : Re: Checkpoints and catch-up burst (heavy back pressure) Hi Arnaud, Thanks for the further feedbacks! For option1: 40min still does not makes sense, which indicates it might take more time to finish checkpoint in your case. I also experienced some scenarios of catching up data to ta

Re: Checkpoints and catch-up burst (heavy back pressure)

.@aliyun.com>> > Envoyé : vendredi 1 mars 2019 04:59 > À : user mailto:user@flink.apache.org>>; LINZ, Arnaud > mailto:al...@bouyguestelecom.fr>> > Objet : Re: Checkpoints and catch-up burst (heavy back pressure) > > Hi Arnaud, > > Thanks for the furt

RE: Checkpoints and catch-up burst (heavy back pressure)

iang -- From:LINZ, Arnaud mailto:al...@bouyguestelecom.fr>> Send Time:2019年3月1日(星期五) 00:46 To:zhijiang mailto:wangzhijiang...@aliyun.com>>; user mailto:user@flink.apache.org>> Subject:RE: Checkpoints and catch-up burst (heavy back pressure) Update : Option 1 does not work.

Re: Checkpoints and catch-up burst (heavy back pressure)

:2019年3月1日(星期五) 00:46 To:zhijiang ; user Subject:RE: Checkpoints and catch-up burst (heavy back pressure) Update : Option 1 does not work. It still fails at the end of the timeout, no matter its value. Should I implement a “bandwidth” management system by using artificial Thread.sleep in the

RE: Checkpoints and catch-up burst (heavy back pressure)

: 'zhijiang' ; user Objet : RE: Checkpoints and catch-up burst (heavy back pressure) Hi Zhihiang, Thanks for your feedback. * I’ll try option 1 ; time out is 4min for now, I’ll switch it to 40min and will let you know. Setting it higher than 40 min does not make much sense since after

RE: Checkpoints and catch-up burst (heavy back pressure)

names, everything else is local to the node). Should I adjust their values nonetheless ? To higher or lower values ? Best, Arnaud De : zhijiang Envoyé : jeudi 28 février 2019 10:58 À : user ; LINZ, Arnaud Objet : Re: Checkpoints and catch-up burst (heavy back pressure) Hi Arnaud, I think there

Re: Checkpoints and catch-up burst (heavy back pressure)

Hi Arnaud, I think there are two key points. First the checkpoint barrier might be emitted delay from source under high backpressure for synchronizing lock. Second the barrier has to be queued in flighting data buffers, so the downstream task has to process all the buffers before barriers to tr

Checkpoints and catch-up burst (heavy back pressure)

Hello, I have a simple streaming app that get data from a source and store it to HDFS using a sink similar to the bucketing file sink. Checkpointing mode is “exactly once”. Everything is fine on a “normal” course as the sink is faster than the source; but when we stop the application for a whil