Details
Description
There are certain operations on the system that depend on the ability to replicate oplog entries in order to make progress. Today, secondaries reading the oplog on their sync sources can be blocked by read ticket exhaustion, potentially creating deadlocks. In order to break this deadlock situation, we can exempt oplog read queries from needing to acquire read tickets.
One good way to determine if a reader is an oplog reader from a downstream node is to use the isReplOplogFetching flag.
Attachments
Issue Links
- is depended on by
-
SERVER-45845 TransactionCoordinator stepUp can deadlock
-
- Closed
-
- is duplicated by
-
-
- Closed
-
-
SERVER-59641 Oplog fetching should not obtain read tickets
-
- Closed
-
- is related to
-
SERVER-44711 Add statistics metrics.repl.network.oplogGetMoresProcessed
-
- Closed
-