Some prerequisite knowledge is documented in SERVER-33159.
Recover to a stable timestamp introduces a case, in crash recovery, where an `_mdb_catalog` collection document may exist without a corresponding WT table. SERVER-33159 describes how that can happen.
SERVER-33159 is a short-term fix that accepts this state can exist. This ticket is to return the happens-before relationship between catalog entries and their on-disk WT tables, in the face of stable timestamp + logging changes for 4.0+ replica set members.
This patch is for glue code to postpone performing the WT table drop until it is sure the collection drop has been persisted in a stable checkpoint.
- duplicates
-
SERVER-38962 The second phase of two-phase drop should occur when the drop becomes both checkpointed and older than the oldest timestamp
- Closed
- is depended on by
-
SERVER-38910 Remove redundant rollback handling on index drops
- Closed
- is related to
-
SERVER-33159 RTT storage recovery from unclean shutdown expects all WT tables to exist
- Closed
-
SERVER-38962 The second phase of two-phase drop should occur when the drop becomes both checkpointed and older than the oldest timestamp
- Closed
-
SERVER-38985 assume all tables exist in storage engine after unclean shutdown
- Closed
- related to
-
SERVER-39007 Switch to use rhel62-large distro for concurrency* tasks on Enterprise RHEL 6.2 (InMemory) builder
- Closed
-
SERVER-29213 Have KVWiredTigerEngine implement StorageEngine::recoverToStableTimestamp
- Closed