It is generally a sign of a healthy system when checkpoints are completing quickly. It would be valuable to log messages periodically if checkpoints are taking a long time to complete.
Those messages would be output via the __wt_msg command and end up in the MongoDB log for MongoDB users.
Ideally we would output a message approximately every 30 seconds. The message should just contain the total elapsed time, and if it's easy/cheap the amount of data written for the checkpoint to this time.
- related to
-
SERVER-38322 add verbose="checkpoint_progress" to WT config
- Closed
-
SERVER-27692 Can diagnostic data capture be stopped later in the shutdown process?
- Closed