mongodb runtime_error ai_generated true

MongoServerError: Transaction aborted due to shard chunk migration in progress

ID: mongodb/transaction-abort-on-shard-migration

Also available as: JSON · Markdown · 中文

85%Fix Rate

84%Confidence

1Evidence

2023-11-05First Seen

Version Compatibility

Version	Status	Introduced	Deprecated	Notes
MongoDB 5.0	active	—	—	—
MongoDB 6.0	active	—	—	—
MongoDB 7.0	active	—	—	—

Root Cause

A multi-document transaction accessed a chunk that was being migrated between shards, causing the balancer to abort the transaction.

generic

中文

多文档事务访问了一个正在分片之间迁移的块，导致平衡器中止了该事务。

Official Documentation

https://www.mongodb.com/docs/manual/core/sharded-cluster-balancer/

Workarounds

90% success Implement a retry loop with exponential backoff to re-run the transaction after a delay: while (true) { try { session.startTransaction(); ... session.commitTransaction(); break; } catch (e) { if (e.code === 251) { sleep(1000 * Math.pow(2, retries)); retries++; } else throw e; } }
```
Implement a retry loop with exponential backoff to re-run the transaction after a delay: while (true) { try { session.startTransaction(); ... session.commitTransaction(); break; } catch (e) { if (e.code === 251) { sleep(1000 * Math.pow(2, retries)); retries++; } else throw e; } }
```
85% success Temporarily disable the balancer during critical transaction windows: sh.stopBalancer(); // run transactions; sh.startBalancer()
```
Temporarily disable the balancer during critical transaction windows: sh.stopBalancer(); // run transactions; sh.startBalancer()
```
75% success Avoid accessing chunks that are being migrated by querying the config.changelog collection to check for recent migrations and routing transactions to stable chunks: use config.changelog to identify migration patterns
```
Avoid accessing chunks that are being migrated by querying the config.changelog collection to check for recent migrations and routing transactions to stable chunks: use config.changelog to identify migration patterns
```

中文步骤

Implement a retry loop with exponential backoff to re-run the transaction after a delay: while (true) { try { session.startTransaction(); ... session.commitTransaction(); break; } catch (e) { if (e.code === 251) { sleep(1000 * Math.pow(2, retries)); retries++; } else throw e; } }

Temporarily disable the balancer during critical transaction windows: sh.stopBalancer(); // run transactions; sh.startBalancer()

Avoid accessing chunks that are being migrated by querying the config.changelog collection to check for recent migrations and routing transactions to stable chunks: use config.changelog to identify migration patterns

Dead Ends

Common approaches that don't work:

50% fail
Stopping the balancer is not a fix for the transaction; it only avoids the symptom but can cause performance issues.
90% fail
The abort is triggered by the migration process, not by timeout.
70% fail
The migration is asynchronous and may take seconds to minutes to complete.