HDDS-11714. resetDeletedBlockRetryCount with --all may fail and can cause long db lock in large cluster #7665
Conversation
This file contains 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
nandakumar131
requested changes
Jan 10, 2025
Comment on lines
123
to
149
| @Override | ||
| public List<DeletedBlocksTransaction> getFailedTransactionsBatch( | ||
| int batchSize, long startTxId) throws IOException { | ||
| List<DeletedBlocksTransaction> failedTXs = new ArrayList<>(); | ||
|
|
||
| lock.lock(); | ||
| try { | ||
| try ( | ||
| TableIterator<Long, ? extends Table.KeyValue<Long, DeletedBlocksTransaction>> iter = | ||
| deletedBlockLogStateManager.getReadOnlyIterator()) { | ||
|
|
||
| iter.seek(startTxId); | ||
|
|
||
| while (iter.hasNext() && failedTXs.size() < batchSize) { | ||
| DeletedBlocksTransaction delTX = iter.next().getValue(); | ||
| if (delTX.getCount() == -1) { | ||
| failedTXs.add(delTX); | ||
| } | ||
| } | ||
| } | ||
| } finally { | ||
| lock.unlock(); | ||
| } | ||
|
|
||
| return failedTXs; | ||
| } | ||
|
|
There was a problem hiding this comment.
Why do we need this additional method? The same thing can be achieved with the existing getFailedTransactions method.
|
|
||
| } while (!batch.isEmpty()); | ||
| } else { | ||
| // Process txIDs provided by the user in batches |
There was a problem hiding this comment.
The user provided list of txIDs reaches SCM via RPC call, so it's ok to process this in single go.
added 3 commits
January 20, 2025 15:20
…ause long db lock in large cluster
aryangupta1998
force-pushed
the
HDDS-11714
branch
from
e2df43e to
88722d6
Compare
nandakumar131
requested changes
Jan 23, 2025
Comment on lines
915
to
918
| AUDIT.logReadFailure( | ||
| buildAuditMessageForFailure( | ||
| SCMAction.GET_FAILED_DELETED_BLOCKS_TRANSACTION, auditMap, ex) | ||
| ); |
There was a problem hiding this comment.
Please avoid unnecessary formatting.
Comment on lines
301
to
302
| // These blocks cannot be found in the container, skip deleting them | ||
| // eventually these TX will success. |
There was a problem hiding this comment.
Please avoid unnecessary formatting.
| @@ -129,7 +129,7 @@ public List<DeletedBlocksTransaction> getFailedTransactions(int count, | |||
| final List<DeletedBlocksTransaction> failedTXs = Lists.newArrayList(); | |||
| try (TableIterator<Long, | |||
| ? extends Table.KeyValue<Long, DeletedBlocksTransaction>> iter = | |||
| deletedBlockLogStateManager.getReadOnlyIterator()) { | |||
| deletedBlockLogStateManager.getReadOnlyIterator()) { | |||
There was a problem hiding this comment.
Please avoid unnecessary formatting.
| @@ -176,18 +176,52 @@ public void incrementCount(List<Long> txIDs) | |||
| */ | |||
| @Override | |||
| public int resetCount(List<Long> txIDs) throws IOException { | |||
There was a problem hiding this comment.
You can simplify the whole logic
public int resetCount(List<Long> txIDs) throws IOException {
if (txIDs != null) {
try {
lock.lock();
transactionStatusManager.resetRetryCount(txIDs);
return deletedBlockLogStateManager.resetRetryCountOfTransactionInDB(
new ArrayList<>(txIDs));
} finally {
lock.unlock();
} catch (Exception e) {
throw new IOException("Error during transaction reset", e);
}
}
final int batchSize = 1000;
int totalProcessed = 0;
long startTxId = 0;
try {
// If txIDs are not provided, fetch all failed transactions in batches
List<DeletedBlocksTransaction> batch;
do {
// Fetch the batch of failed transactions
batch = getFailedTransactions(batchSize, startTxId);
List<Long> batchTxIDs = batch.stream().map(DeletedBlocksTransaction::getTxID).collect(Collectors.toList());
totalProcessed += resetCount(batchTxIDs);
// Update startTxId to continue from the last processed transaction in the next iteration
startTxId = batch.get(batch.size() - 1).getTxID() + 1;
} while (!batch.isEmpty());
}
retrun totalProcessed;
}
added 2 commits
January 24, 2025 10:24
|
@aryangupta1998 the test failure seems related to this change, can you take a look at it? |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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 changes were proposed in this pull request?
In case of resetDeletedBlockRetryCount with --all option, scm takes lock and tries to get all the transaction with max retry and then updates DB with 0 count. In some large scale env this count can be huge which can lead to multiple problem.
i) Lock can lead to block all other normal operation.
ii) Since message is passed through ratis, which will fail because of size.
Instead of doing like above we should do this operation in batches to avoid long lock and ratis message size failure.
What is the link to the Apache JIRA
https://issues.apache.org/jira/browse/HDDS-11714
How was this patch tested?
Tested Manually.