Highh memory commit during cleanup on Exchange RRS feed

  • Question

  • We have an Exchange 2010 installation with around 14000 mailboxes. Running DAG with two copies of each database with active mailboxes distributed among all 8 mailbox servers. We are running single role servers.

    We are experiencing an issue where the committed RAM for store.exe grows very high and grows to more than double than normal. This results in heavy swapping (some 300 hard faults/sec), slow performance and results in unhealthy databases.

    This does not occur during normal hours – but when it happens it can be many hours before store.exe and the server returns to normal.

    We are suspecting two tasks that can result in this. One is full software backup with Tivoli TSM. The other is exchange database maintenance.

    We are seeing in the event log that database maintenance takes around 10 minutes – cleanup of items and mailboxes past retention date, cleanup of content index failed items. All finishes in around ten minutes. But shortly after this we get the first warning “A significant portion of the database buffer cache has been written out to the system paging file” followed during the next 20 minutes buy several others where the Current Total Percent Resident drops to below 30%. This even though the tasks seems to be finished.

    It will then take several hours before the server returns to normal.

    Is anybody else seeing a heavy growth in committed RAM after the maintenance tasks? Any theories why this happens, and what we can do to avoid this?

    Thursday, June 27, 2013 6:24 PM