Description
Currently, we do creation for partition and index files during PME for starting caches at the beginning of org.apache.ignite.internal.processors.cache.persistence.GridCacheDatabaseSharedManager#readCheckpointAndRestoreMemory method.
We should avoid this because sometimes it took a long time as we perform writing to disk.
If the cache was registered during PME we should initialize page store lazy or asynchronously.
Attachments
Issue Links
- is related to
-
IGNITE-7196 Exchange can stuck and wait while new node restoring state from disk and starting caches
- Resolved
- links to