klcopp commented on a change in pull request #2971:
URL: https://github.com/apache/hive/pull/2971#discussion_r792628396
##########
File path: ql/src/java/org/apache/hadoop/hive/ql/txn/compactor/Cleaner.java
##########
@@ -434,8 +437,18 @@ private boolean removeFiles(String location,
ValidWriteIdList writeIdList, Compa
return success;
}
- private boolean hasDataBelowWatermark(FileSystem fs, Path path, long
highWatermark) throws IOException {
- FileStatus[] children = fs.listStatus(path);
+ private boolean hasDataBelowWatermark(AcidDirectory acidDir, FileSystem fs,
Path path, long highWatermark)
+ throws IOException {
+ Set<Path> acidPaths = new HashSet<>();
+ for (ParsedDelta delta : acidDir.getCurrentDirectories()) {
+ acidPaths.add(delta.getPath());
+ }
+ if (acidDir.getBaseDirectory() != null) {
+ acidPaths.add(acidDir.getBaseDirectory());
+ }
+ FileStatus[] children = fs.listStatus(path, p -> {
+ return !acidPaths.contains(p);
+ });
for (FileStatus child : children) {
if (isFileBelowWatermark(child, highWatermark)) {
Review comment:
1.
> I believe that in case there are files in the dir they already should be
in the obsolete list
Not necessarily, because the AcidDirectory the Cleaner uses is computed
based on an older txnId (cleanerWaterMark), so there is a chance its obsolete
list does not contain files that should be cleaned up eventually, which is what
this method is supposed to figure out. (Right?)
@deniskuzZ please correct me if I'm wrong about this since I know there have
been recent changes to this logic
2. I meant that if the table dir contains:
- delta_5_5
- delta_1_5_v100 (minor compacted)
Then the cleaner should eventually remove delta_5_5, so there will be files
to remove later, when the cleanerWaterMark is high enough
--
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]
