apache/pulsar

same pulsar-admin should seamlessly work with both brokers and function workers in a distributed environment

Open

#6,425 opened on Feb 26, 2020

View on GitHub
 (2 comments) (2 reactions) (0 assignees)Java (13,697 stars) (3,485 forks)batch import
area/functiongood first issuehelp wantedtype/feature

Description

I am running Function worker as a separate machine, and pulsar-admin gives this error:

# bin/pulsar-admin functions create   --py c.py   --classname router.RoutingFunction   --tenant c1-d1   --namespace emsaas   --name test1   --inputs persistent://c1-d1/emsaas/in
Function worker service is not done initializing. Please try again in a little while.
Reason: HTTP 503 Service Unavailable

client.conf points to the broker. Changing webServiceUrl to point to the function worker fixes the issue. This is not intuitive and the error message doesn't help. I would expect pulsar-admin to get the Function worker address from the broker somehow and work transparently without the need to use different client.conf's.

Relevant broker log: 172.32.2.235 is the address of the Function worker, and the log that fails it seems to connect to 127.0.0.1.

Feb 23 13:17:09 localhost pulsar: 13:17:09.892 [pulsar-web-30-15] INFO  org.eclipse.jetty.server.RequestLog - 172.32.2.235 - - [23/Feb/2020:13:17:09 +0000] "GET /admin/v2/persistent/public/functions/coordinate/stats HTTP/1.1" 200 901 "-" "Pulsar-Java-v2.5.0" 1
Feb 23 13:17:39 localhost pulsar: 13:17:39.879 [pulsar-web-30-8] INFO  org.eclipse.jetty.server.RequestLog - 172.32.2.235 - - [23/Feb/2020:13:17:39 +0000] "GET /admin/v2/persistent/public/functions/coordinate/stats HTTP/1.1" 200 901 "-" "Pulsar-Java-v2.5.0" 3
Feb 23 13:17:41 localhost pulsar: 13:17:41.644 [pulsar-web-30-4] INFO  org.eclipse.jetty.server.RequestLog - 127.0.0.1 - - [23/Feb/2020:13:17:41 +0000] "POST /admin/v3/functions/c1-d1/emsaas/test1 HTTP/1.1" 503 98 "-" "Pulsar-Java-v2.5.0" 4
Feb 23 13:17:54 localhost kernel: ena: Feature 27 isn't supported
Feb 23 13:18:09 localhost pulsar: 13:18:09.879 [pulsar-web-30-6] INFO  org.eclipse.jetty.server.RequestLog - 172.32.2.235 - - [23/Feb/2020:13:18:09 +0000] "GET /admin/v2/persistent/public/functions/coordinate/stats HTTP/1.1" 200 901 "-" "Pulsar-Java-v2.5.0" 1

Contributor guide

Tech stack
javashell
Domain
backendapi
Issue type
feature
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
4
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
3-5 days
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
stale
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
mostly clear
Prerequisites
Apache Pulsar basicsJava programmingHTTP/REST
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
25
Research direction
The issue requires modifying the pulsar admin CLI to automatically discover the Function Worker address from the broker. Currently, the client uses a fixed webServiceUrl. Key files to investigate include the Admin client (e.g., PulsarAdmin.java) and the broker's admin API for function worker discovery. The error message 'Function worker service is not done initializing' from the broker suggests a 503 when the broker tries to forward to a local function worker. Look at how the Admin client handles function operations and how it could query the broker for the function worker endpoint (e.g., via /admin/v2/worker or similar). A possible approach is to add a redirect or discovery mechanism in the client.