[
https://issues.apache.org/jira/browse/HIVE-25492?focusedWorklogId=751661&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-751661
]
ASF GitHub Bot logged work on HIVE-25492:
-----------------------------------------
Author: ASF GitHub Bot
Created on: 01/Apr/22 16:19
Start Date: 01/Apr/22 16:19
Worklog Time Spent: 10m
Work Description: asinkovits commented on a change in pull request #3157:
URL: https://github.com/apache/hive/pull/3157#discussion_r840739714
##########
File path: ql/src/java/org/apache/hadoop/hive/ql/io/AcidUtils.java
##########
@@ -1480,6 +1471,40 @@ private static ValidTxnList
getValidTxnList(Configuration conf) {
return validTxnList;
}
+
+ /**
+ * In case of the cleaner, we don't need to go into file level, it is enough
to collect base/delta/deletedelta directories.
+ *
+ * @param fs the filesystem used for the directory lookup
+ * @param path the path of the table or partition needs to be cleaned
+ * @return The listed directory snapshot needs to be checked for cleaning
+ * @throws IOException on filesystem errors
+ */
+ public static Map<Path, HdfsDirSnapshot> getHdfsDirSnapshotsForCleaner(final
FileSystem fs, final Path path)
+ throws IOException {
+ Map<Path, HdfsDirSnapshot> dirToSnapshots = new HashMap<>();
+ // depth first search
+ Deque<RemoteIterator<FileStatus>> stack = new ArrayDeque<>();
+ stack.push(fs.listStatusIterator(path));
+ while (!stack.isEmpty()) {
+ RemoteIterator<FileStatus> itr = stack.pop();
+ while (itr.hasNext()) {
+ FileStatus fStatus = itr.next();
+ Path fPath = fStatus.getPath();
+ if (acidHiddenFileFilter.accept(fPath)) {
+ if (deltaFileFilter.accept(fPath) ||
+ baseFileFilter.accept(fPath) ||
+ deleteEventDeltaDirFilter.accept(fPath)) {
+ dirToSnapshots.put(fPath, new HdfsDirSnapshotImpl(fPath));
+ } else {
Review comment:
Yeah, I've reiterated this part as there were issues around nonacid-acid
conversion.
--
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]
Issue Time Tracking
-------------------
Worklog Id: (was: 751661)
Time Spent: 1h 40m (was: 1.5h)
> Major query-based compaction is skipped if partition is empty
> -------------------------------------------------------------
>
> Key: HIVE-25492
> URL: https://issues.apache.org/jira/browse/HIVE-25492
> Project: Hive
> Issue Type: Bug
> Reporter: Karen Coppage
> Assignee: Antal Sinkovits
> Priority: Major
> Labels: pull-request-available
> Time Spent: 1h 40m
> Remaining Estimate: 0h
>
> Currently if the result of query-based compaction is an empty base, delta, or
> delete delta, the empty directory is deleted.
> This is because of minor compaction – if there are only deltas to compact,
> then no compacted delete delta should be created (only a compacted delta). In
> the same way, if there are only delete deltas to compact, then no compacted
> delta should be created (only a compacted delete delta).
> There is an issue with major compaction. If all the data in the partition has
> been deleted, then we should get an empty base directory after compaction.
> Instead, the empty base directory is deleted because it's empty and
> compaction claims to succeed but we end up with the same deltas/delete deltas
> we started with – basically compaction does not run.
> Where to start? MajorQueryCompactor#commitCompaction
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
