Description
Stopping node during checkpoint leads to binary memory recovery after node start.
When binary memory is restored node performs checkpoint that fixes the consistent state of the page memory.
It happens there
org.apache.ignite.internal.processors.cache.persistence.GridCacheDatabaseSharedManager#finalizeCheckpointOnRecovery
Looking at the implementation of this method we can notice that it performs finalization in 1 thread, which is not optimal. This process can be speed-up using parallelization of collecting checkpoint pages like in regular checkpoints.
Attachments
Issue Links
- incorporates
-
IGNITE-10508 Need to support the new checkpoint feature, checkpint should not wait for the previous operation to complete
- Open
- links to