pulsar standalone docker image should support initialization of some kind

[flowchartsman]

This would be very useful for development and integration testing. Currently have to run a parallel container calling pulsar-admin in a loop to wait on the relevant endpoints to be up and spin up the tenants/namespaces/policies that I need to simulate my deployment environment.

Perhaps this could be read from a file at a known location and then spun up in place fo the sample tenant/namespaces if it's found?

[flowchartsman]

My suggestion would be something like the following:

1. before spinning up the admin endpoint, look for a file in a known location (following mongo's example, this would be something like /docker-entrypoint-init.d/pulsar-init
2. If this file is found, process it
3. otherwise, proceed with sample tenant/namespace creation

As for the format of the file, one option would be to simply treat it as a newline-separated list of arguments to pulsar-admin invocations:

tenants create test
namespaces create test/admin
namespaces create test/news
namespaces create test/emails
namespaces set-retention --size-1 --time 3d test/news
namespaces set-publish-rate -m 10 test/emails

[tisonkun]

Moved to the General Discussions forum. I like the idea but it requires a design (PIP) and volunteers working on it. Discussions forum is sorted default by updated and gives this thread more attention.

Perhaps .hiverc to Hive. For your suggestion command pattern, the problem is which bin script to execute, shall we support run bin/pulsar ... also?

[flowchartsman]

shall we support run bin/pulsar ... also?

Not for the use-case I had in mind, which was for pulsar, specifically in standalone mode. Actually, even more specifically, it was targeted at the pulsar-standalone image to help out in development and testing. Since I originally wrote this issue, it seems the dedicated pulsar-standalone image has fallen behind, so it makes it tricker to implement, since you'd either need to have a dedicated image, or build it into pulsar directly, and that could be problematic if misused.

Since then, I've also found a technique that works "well enough", by using a script which calls curl on the metrics endpoint in a loop, then setting this script as the as the container "health check":

#!/bin/bash
cd /pulsar/bin
# Wait for pulsar standalone to be initialized
until curl http://localhost:8080/metrics > /dev/null 2>&1 ; do sleep 1; done
./pulsar-admin tenants create -c standalone mytenant
./pulsar-admin namespaces create mytenant/mynamespace
<etc>

So I guess my question at this point would be: is it useful to provide this functionality as described, or is there a compelling use-case for calling /pulsar/bin/pulsar as well? I'm not sure, by I still think this would be handy, even if it's config-file driven.