FlushToDiskOnStop | Performance tuning

Limitation on invalidation when server is stopped

When a server is stopped, it can no longer receive and process any cache replication requests.

There is a limitation on invalidation when a server is stopped.

Consider this scenario. FlushToDiskOnStop is enabled, and upon a server stop, in-memory and disk cache are preserved on the file system.

One server in the cluster is stopped, but the other ones are running and processing invalidation requests. Often the invalidated cache entry also exists in the disk cache of the server that is stopped. It will not have the ability to remove this cache entry from its disk cache, and therefore when this cluster member starts back up, its disk cache will contain outdated cache entries.

Because of this limitation, in order to keep the cache content consistent, you may wish to disable FlushToDiskOnStop in a clustered environment. The downside to this is that when a server starts up, its cache content must be rebuilt. As such, the initial response time from this server will be slow.

xxxx