pjfanning opened a new pull request, #1918:
URL: https://github.com/apache/pekko-connectors/pull/1918
Closes out #1912 (`hbase-common`/`hbase-shaded-client` 2.6.6 -> 3.0.0),
which fails every hbase test.
### Why the 3.0.0 bump fails
hbase 3.0.0 flipped the default connection registry. From
`ConnectionRegistryFactory` in each client jar:
| hbase-client | default for `hbase.client.registry.impl` |
| --- | --- |
| 2.6.6 | `ZKConnectionRegistry` |
| 3.0.0 | `RpcConnectionRegistry` |
Our tests build their config from a bare `HBaseConfiguration.create()`
(`hbase/src/test/scala/docs/scaladsl/HBaseStageSpec.scala`) and there is no
`hbase-site.xml` under `hbase/src/test/resources`, so a 3.x client no longer
goes to ZooKeeper on 2181. It falls back to the master at `<hostname>:16000`
and calls `ClientMetaService`.
The test server is `harisekhon/hbase:2.1`, which predates
`MasterRegistry`/`ClientMetaService` (added in 2.3), so the master just closes
the socket. All 7 tests fail the same way at connection setup:
```
MasterRegistryFetchException: Exception making rpc to masters [...:16000]
Cause: RetriesExhaustedException: Failed contacting masters after 1
attempts.
ConnectionClosedException: Call to address=...:16000 failed on local
exception
```
with this in the debug log:
```
ClusterIdFetcher: Failed to get connection registry info, should be an old
server,
fallback to use null cluster id
```
### Why pin rather than fix
`harisekhon/hbase` has no tag newer than `2.1` and there is no official
`apache/hbase` image, so moving the test server to 3.x is not a small change.
And a 3.x client against a 2.1 server is not a combination hbase supports, so
forcing `ZKConnectionRegistry` in the tests would only paper over a real
behaviour change for users.
If we later stand up a real hbase 3 test server, drop this pin.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
