-
Type: Task
-
Resolution: Fixed
-
Priority: Major - P3
-
Affects Version/s: None
-
Labels:
-
0.8
Documentation Request Summary:
This ticket added a new section to the serverStatus response called "transactions" that contains 3 new fields:
- "retriedCommandsCount"
- The total number of received write commands with statements that have already been executed, across all sessions
- "retriedStatementsCount"
- The total number of received statements that have already been executed, across all sessions
- "transactionsCollectionWriteCount"
- The total number of writes to the config.transactions collection, triggered when a new retryable write statement is committed. For updates and deletes there is one write per statement, but there is only one per batch of documents inserted, except when there is a failure that leads to each document being inserted separately.
"retriedCommandsCount" and "retriedStatementsCount" both ignore any retryable writes that may happen internally as part of a chunk migration, but "transactionsCollectionWriteCount" does include writes to a server's config.transactions collection that occur as part of a migration.
All three are NumberLongs.
Scope of changes:
- serverStatus (backport to 3.6)
- retryable writes (backport to 3.6)
Impact to other docs outside of this product:
none
MVP:
Resources:
Engineering Ticket Description:
This ticket is a placeholder to figure out between ourselves and with TSE what supportability and diagnosability enhancements we might need to include for retryable writes.
Things that come to mind include:
- FTDC metrics for number of active sessions
- Metrics for the CPU (and if possible I/O) overhead of maintaining the retryability metadata
- Metrics for number of writes which have been retried
- If we end up with a cache, metrics for hit-rate
- documents
-
SERVER-28923 Add diagnosability and supportability features for retryable writes
- Closed