smengcl opened a new pull request, #8878: URL: https://github.com/apache/hadoop/pull/8878
Generated-by: Claude Code (Opus 5.5) JIRA: https://issues.apache.org/jira/browse/YARN-12032 ### Description of PR `RMNodeImpl` built the node RPC and HTTP addresses as `hostName + ":" + port`. When a NodeManager host has no reverse DNS, its NodeId host is a numeric IPv6 literal, so the HTTP address became `2001:db8:0:0:0:0:0:21:8042`. That value is `Container.getNodeHttpAddress()`, `NodeReport.getHttpAddress()` and the REST `nodeHTTPAddress`. `NetUtils.createSocketAddr` rejects it, so the MR AM fails when a container launches, and `new URL("http://"; + address)` throws, which breaks node and log links in the web UI and the log redirect. This change: - Builds both addresses in `RMNodeImpl` with `NetUtils.getHostPortString`, which brackets IPv6 literals. Host names, IPv4 literals and a null host give the same string as before. - `TaskImpl` took the shuffle host with `getNodeHttpAddress().split(":")[0]`, which would give `[2001` for a bracketed address. It now parses the address with `NetUtils.createSocketAddrUnresolved`, the same parser `TaskAttemptImpl` already applies to this value when the container launches, and rebuilds `host:port` with `getHostPortString`. - `TaskAttemptImpl.recover` rebuilds the node HTTP address from job history the same way, so the recovered value stays parseable by the new `TaskImpl` code. For IPv6 this path is still blocked one line earlier by `NodeId.fromString`, which YARN-12033 covers. The consumers that split this value on `:` were checked. The others only display it, prefix a scheme, or parse it with `NetUtils.createSocketAddr`, which accepts brackets. ### How was this patch tested? - `TestRMNodeTransitions#testNodeAddressesBracketIPv6Host` (new): fails without the fix with `expected: <[2001:db8:0:0:0:0:0:21]:45454> but was: <2001:db8:0:0:0:0:0:21:45454>`. It also checks that host name and IPv4 addresses are unchanged. - `TestTaskImpl#testMapOutputServerAddress` (new): fails without the `TaskImpl` change with `http://[2001:13562` and checks the host name case. - `TestRMNodeTransitions`, `TestRMWebServicesNodes`, `TestResourceTrackerService`, `TestTaskImpl`, `TestTaskAttempt` and `TestRecovery` pass. - Not run on a cluster. ### For code changes: - [x] Does the title or this PR starts with the corresponding JIRA issue id (e.g. 'HADOOP-17799. Your PR title ...')? - [ ] Object storage: have the integration tests been executed and the endpoint declared according to the connector-specific documentation? - [ ] If adding new dependencies to the code, are these dependencies licensed in a way that is compatible for inclusion under [ASF 2.0](http://www.apache.org/legal/resolved.html#category-a)? - [ ] If applicable, have you updated the `LICENSE`, `LICENSE-binary`, `NOTICE-binary` files? 🤖 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]
