mongodb/mongo
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

SERVER-101673 Documentation Updates (#40293)

Browse Source 
GitOrigin-RevId: a1e51012d36abf04ac231cbcb95938b7242ff2e4
This commit is contained in:
clarissecheah 2025-08-26 11:06:21 +10:00 committed by MongoDB Bot
parent ed6a5c217e
commit 85f50fb465
1 changed files with 6 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
src/mongo/db/storage/wiredtiger/README.md
View File

@ -216,6 +216,12 @@ A new truncate marker is created when the in-progress marker segment contains mo
bytes needed to complete the segment; and the oldest truncate marker's oplog is deleted when the
oplog size exceeds its cap size setting.
Oplog sampling operates in two modes: asynchronous and synchronous. By default, it uses asynchronous mode, where oplog sampling and initial marker generation happen in the background as part of the OplogCapMaintainer thread. This setup ensures that startup is not blocked and oplog reads and writes are available during tartup. Until the initial marker generation is complete, no new truncate markers can be created for these new oplog writes.
Asynchronous mode offers better performance, allowing faster startups and restarts while improving overall node availability. One potential trade-off to consider is the possible increased disk usage.
You can switch between asynchronous and synchronous modes by adjusting the OplogSamplingAsyncEnabled server parameter.
Oplog sampling and marker generation is skipped when using `--restore` or `--magicRestore`.
## Special Timestamps That Will Not Be Truncated