This sounds about right to me. When I was working on getting flink running this was one of the biggest pain points, the upload/download process is painfully slow. We had implemented a jar cache in flink to improve this (and stopped using fat jars) https://github.com/twitter-forks/flink/commit/01f26ab7124b722ec767c8e01d10395afbbb0dc5 but I don't think it ever made it upstream.
On Tue, May 7, 2024 at 10:32 AM Balogh, György <bog...@ultinous.com> wrote: > Hi, > We are switching from embedded flink runner to filink server. The embedded > flink runner works fine, jobs start instantly. However when we use the > flink server we are experiencing about 30 seconds of idle time (no > significant cpu, io load) before we see the job in the flink. We are > targeting at most 2 seconds overhead to run ad-hoc query pipelines. > We did some investigation and saw that the fat jar is about 150MB that was > sent from the beam to the flink cluster. Still this does not justify the 30 > sec. We are trying to deploy dependencies to the flink workers and decrease > the jar size. It's not clear at the moment what causes this delay. > > Flink version: 1.16 > Beam version: 2.55.1 > > Thank you, > Gyorgy > > -- > > György Balogh > CTO > E gyorgy.bal...@ultinous.com <zsolt.sala...@ultinous.com> > M +36 30 270 8342 <+36%2030%20270%208342> > A HU, 1117 Budapest, Budafoki út 209. > W www.ultinous.com >