Re: Stuck in Fast Recovery?

Stephen Walmsley wrote: > > Many thanks for the hints June and Paul. > > Eventually, after 7.5 million reads with no write activity I gave up. I > did try a shutdown followed by a restart as Paul suggested, but the engine > just continued the same pattern of activity. I have no idea whether it > would finally have gone online but I couldn't afford to wait and see. The > restoration is in progress... Stephen, What you experienced is actually a rare and freak occurrance. Even with unbuffered logging the current log buffer is at risk in case of a system crash. There are a few critical event records that if not found in the ondisk log file can cause the engine to go into permanent recovery. If such critical record was required and in the log buffer about to be written and the records that caused fast recovery to expect such a record type had already been flushed in a previous buffer, voila! Usually tech support can log in and wipe out the unmatched record(s) at the end of the last log file and the engine recovers fine. The permanent solution is problematic as there are two schools of thought on it (both mine) 1) decrease the size of the Logical log buffer so less data is at risk (always a good thing but a crap shoot) or 2) Increase the size of the Logical Log buffer so that it is less likely that a critical record pair can be partially written (also a crap shoot). I tend to lean toward option 1) myself. Art S. Kagel