pulsar-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Apache Pulsar Slack" <apache.pulsar.sl...@gmail.com>
Subject Slack digest for #general - 2018-12-22
Date Sat, 22 Dec 2018 09:11:02 GMT
2018-12-21 13:25:05 UTC - borlandor: @borlandor has joined the channel
2018-12-21 16:21:59 UTC - Haymon Edmur: @Haymon Edmur has joined the channel
2018-12-21 16:40:34 UTC - Shalin: Does creating a pulsar function in a broker create instances
of pulsar functions in that broker or all brokers in the cluster?
2018-12-21 16:41:54 UTC - Grant Wu: I don’t know what you mean by “in that broker” or
“all brokers”
2018-12-21 16:42:09 UTC - Grant Wu: <http://pulsar.apache.org/docs/en/functions-deploying/>
does this help?  In general I don’t think how you deploy it impacts the behavior, just the
2018-12-21 16:42:41 UTC - Grant Wu: &gt; the function code will be uploaded to a Pulsar
broker and run alongside the broker rather than in your local environment.
seems to imply just one broker, but not clear
2018-12-21 16:46:22 UTC - Shalin: Right, I was concerned about the performance part. Like
if `broker B` is serving `topic 1` and the function `test` reads from `topic 1` but is deployed
on `broker A`, it might not be ideal.
thinking_face : Grant Wu
2018-12-21 16:48:05 UTC - Shalin: But then again a function would be interacting with multiple
topics served in different brokers, so not sure there could be locality aware deployment of
2018-12-21 17:02:03 UTC - Shalin: @Grant Wu: Did that make sense? or is that not how pulsar
functions work ?
2018-12-21 17:02:11 UTC - Grant Wu: Yeah it made sense
2018-12-21 17:02:24 UTC - Grant Wu: You’ll have to wait for a response from a Pulsar dev
2018-12-21 17:02:52 UTC - Shalin: Lol, gotcha. :thumbsup:
2018-12-21 18:35:46 UTC - Sanjeev Kulkarni: @Shalin pulsar functions running in a pulsar cluster
are distributed across all function workers in that cluster. The logic of which function instance
goes to which worker is pluggable. The current logic just does a round robin assignment and
thus is essentially random. However the mechanism is meant to be pluggable and we can have
other assignment logic like scheduling functions closest to the broker that serves input/output
topic. This is an area of active development 
2018-12-21 20:06:34 UTC - Michael Weedon: @Michael Weedon has joined the channel
2018-12-21 20:43:04 UTC - Daniel Root: I'm trying to reduce the number of log messages produced
by the standalone pulsar docker container.  I've modified the log4j2.yaml file  but can't
figure out how to load the changes.  If I restart the container my modifications are of course
gone.  How do I load these changes?  Restart the cluster?  If so how?  I do have access to
the admin-cli.
View raw message