Closed Ycallaer closed 2 months ago
Hi, I was able to solve it. The blogs I listed were a bit contradictory as to where the JARs needed to be.
I refactored the code to have all the JARs in druid-protobuf-extensions and now I am able to submit the job. The UI no longer throws an error and from the supervisor I can see the schema registry definition is kept intact.
Hi, As part of a POC we are trying to ingest kafka protobuf messages, the schema is stored in schema registry. We are following these 2 blogs:
Since this is a POC, no security is present on the kafka cluster ( we are using Confluent CFK, the demo repo that they made available).
To run druid we are using https://github.com/datainfrahq/druid-operator.
The ingestion configuration we are submitting from the UI is the following
When this is submitted, the following stacktrace is generated from the UI
If we just provide topic and broker we get a successful connection, but we get binary data. So we know it is not a network / connectivity issue. The moment the schema registry part is added is when we get errors.
Affected Version
I have tested this with the image of Apache druid 25.0.0 and 28.0.1 and 29.0.1
Description
Please include as much detailed information about the problem as possible.
Cluster size: The cluster definition we are using is the tiny cluster from https://github.com/datainfrahq/druid-operator/blob/master/examples/tiny-cluster.yaml
From the e2e example dir , I modified the file https://github.com/datainfrahq/druid-operator/blob/master/e2e/configs/druid-cr.yaml , and added in the common runtime the extension druid-protobuf-extensions Additionally , through a side container I have added the following JARs
If I exec into the pod, I can see that the JARs are present on the system.
If we change the keyword
inputFormat
toparser
(the old syntax way), the job submits, but the configuration part of schema registry dissapears and the job in the end fails stating thatinputFormat
block is missing.I had already reached out in the slack channel, and they requested me to log an incident.
Additionally, when these jobs get triggered I can see that the following libraries being loaded but not the 3 libs that I add through the side container. Not sure if that is relevant or not.