Uh oh!
There was an error while loading. Please reload this page.
[fix](be) Include HDFS connection in file handle cache key - #63516
Merged
Conversation
### What problem does this PR solve? Issue Number: None Related PR: None Problem Summary: HDFS file handles were cached only by path and mtime, so a later query using a different hdfsFS authentication context could reuse a handle opened from another context when the same file path and mtime matched. ### Release note None ### Check List (For Author) - Test: Manual test - `build-support/clang-format.sh be/src/io/fs/file_handle_cache.cpp be/src/io/fs/file_handle_cache.h be/test/io/fs/file_handle_cache_test.cpp` - `git diff --cached --check` - BE UT not run locally per request - Behavior changed: No - Does this need documentation: No
hello-stephen
commented
May 22, 2026
Contributor
Thank you for your contribution to Apache Doris. Please clearly describe your PR:
|
suxiaogang223
commented
May 22, 2026
MemberAuthor
run be-ut |
suxiaogang223
commented
May 22, 2026
MemberAuthor
run buildall |
suxiaogang223
marked this pull request as ready for review
May 22, 2026 06:05
hello-stephen
commented
May 22, 2026
Contributor
TPC-H: Total hot run time: 31854 ms |
hello-stephen
commented
May 22, 2026
Contributor
TPC-DS: Total hot run time: 169339 ms |
hello-stephen
commented
May 22, 2026
Contributor
BE UT Coverage ReportIncrement line coverage Increment coverage report
|
hello-stephen
commented
May 22, 2026
Contributor
BE Regression && UT Coverage ReportIncrement line coverage Increment coverage report
|
Gabriel39
approved these changes
May 25, 2026
Contributor
PR approved by at least one committer and no changes requested. |
Contributor
PR approved by anyone and no changes requested. |
Gabriel39
commented
May 25, 2026
Contributor
/review |
Contributor
There was a problem hiding this comment.
Automated review result: no blocking issues found.
Critical checkpoint conclusions:
- Goal and proof: The PR addresses HDFS file-handle cache reuse across different hdfsFS/authentication contexts by including hdfsFS in both the cache key and partition hash. The added BE unit test verifies key equality changes for hdfsFS, path, and mtime.
- Scope and clarity: The change is small and focused on the file-handle cache keying logic.
- Concurrency: The existing LruMultiCache locking model is preserved; the change only affects key construction and partition selection before acquiring the partition cache lock. No new shared mutable state or lock ordering is introduced.
- Lifecycle/static initialization: The existing function-local static cache in HdfsFileReader is unchanged. The hdfsFS handle is already owned by HdfsMgr/HdfsHandler, and this PR does not introduce new static/global objects.
- Configuration/compatibility: No new configuration, protocol, storage format, or FE-BE compatibility concerns.
- Parallel paths: The only HDFS reader cache entry point goes through HdfsFileReader::create -> get_file -> FileHandleCache::get_file_handle, so the changed path covers the relevant use.
- Tests: New BE unit test is included under be/test and should be picked up by the existing GLOB_RECURSE UT_FILES. I also ran
git diff --check cbe23a31d8d37f4e25bcc40e945d6559299903b7 origin/pr/63516with no whitespace errors. Full BE UT was not run in this review environment. - Observability: No additional observability appears necessary for this small keying fix; existing VLOG close/unbuffer logging remains intact.
- Data correctness/transactions/persistence: Not applicable; no transaction, version, MoW, or persistence behavior is modified.
- Performance: The additional pointer hash/key comparison is negligible and avoids incorrect cross-context reuse.
User focus points: no additional user-provided review focus was specified.
Uh oh!
There was an error while loading. Please reload this page.
github-actionsBot
pushed a commit
that referenced
this pull request
May 26, 2026
HDFS file handles were cached only by file path and mtime. In a pipeline that runs multiple Hive catalog cases against the same HDFS path shape, a later query can use a different `hdfsFS` authentication context while still matching the old cache key. That can reuse a handle opened under another context and lead to HDFS auth failures such as `SIMPLE authentication is not enabled. Available:[TOKEN, KERBEROS]`. This PR includes the `hdfsFS` identity in the file handle cache key and partition hash, so cached handles are only reused within the same filesystem/authentication context.
This was referenced May 26, 2026
github-actionsBot
pushed a commit
that referenced
this pull request
May 26, 2026
HDFS file handles were cached only by file path and mtime. In a pipeline that runs multiple Hive catalog cases against the same HDFS path shape, a later query can use a different `hdfsFS` authentication context while still matching the old cache key. That can reuse a handle opened under another context and lead to HDFS auth failures such as `SIMPLE authentication is not enabled. Available:[TOKEN, KERBEROS]`. This PR includes the `hdfsFS` identity in the file handle cache key and partition hash, so cached handles are only reused within the same filesystem/authentication context.
morningman pushed a commit
that referenced
this pull request
May 28, 2026
yiguolei pushed a commit
that referenced
this pull request
May 29, 2026
Closed
zhaorongsheng pushed a commit
to zhaorongsheng/doris
that referenced
this pull request
Jun 4, 2026
) HDFS file handles were cached only by file path and mtime. In a pipeline that runs multiple Hive catalog cases against the same HDFS path shape, a later query can use a different `hdfsFS` authentication context while still matching the old cache key. That can reuse a handle opened under another context and lead to HDFS auth failures such as `SIMPLE authentication is not enabled. Available:[TOKEN, KERBEROS]`. This PR includes the `hdfsFS` identity in the file handle cache key and partition hash, so cached handles are only reused within the same filesystem/authentication context.
Closed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for freeto join this conversation on GitHub.
Already have an account?
Sign in to comment
5 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What problem does this PR solve?
HDFS file handles were cached only by file path and mtime. In a pipeline that runs multiple Hive catalog cases against the same HDFS path shape, a later query can use a different
hdfsFSauthentication context while still matching the old cache key. That can reuse a handle opened under another context and lead to HDFS auth failures such asSIMPLE authentication is not enabled. Available:[TOKEN, KERBEROS].This PR includes the
hdfsFSidentity in the file handle cache key and partition hash, so cached handles are only reused within the same filesystem/authentication context.Issue Number: None
Related PR: None
Release note
None
Check List (For Author)
build-support/clang-format.sh be/src/io/fs/file_handle_cache.cpp be/src/io/fs/file_handle_cache.h be/test/io/fs/file_handle_cache_test.cpp