scroodj,


A "Bufunhash" message usually means that a cache management error occurred. It seems that a desynchronization occurred between the page number present in cache and the page number on disk.

As seen, the message indicates 'no action is necessary'. "Bufunhash" errors are treated as transient memory corruption messages for the buffers. This category of error does not raise a server error, but rather a message will be written into the errorlog when unhashing a buffer results into a page mismatch error.

In fact this message may not be as harmless as it looks like at first.

1) The reasons why this message is printed can be very internals to ASE, few of these reasons can be :
- If the objid on the page just read in is 0 (i.e an uninitialized page)
- The objid on the page is not matching with the objid on the SDES (System Descriptor)
- An aborted APF (Asynchronous PreFecth) read due to lack of resources
.....

But most of the time the "Bufunhash" message is produced during errors like 605, 692 ,695, 697, 834.... In such situation, the root cause can be diagnosed with the printed error number, and most of the time these errors are due to a database corruption or a hardware problems.

If no other errors occur with the message "Bufunhash: Buffer xxx from database id x has page number x in the page header and page number y in the buffer header. Buffer will be unhashed", then this may only be an in-memory problem -- a transient error.

We recommend to run 'dbcc checkdb' and 'dbcc checkalloc' on the database to make sure there are no other error messages.

More information about message "Buffer will be unhashed": When a page is read into memory from the disk, a buffer header is assigned to it. The header tracks the usage and the contents of the page image and together the page and the buffer header define a buffer, it's these buffers themselves that are grouped together to form MASSes (Memory Address Space Segments) and this is saved into special(hash)tables, so basically it means a buffer is removed from a hash table.

3) We strongly advise to check the OS log for any errors being reported at the hardware level. This message is not harmful. It could be an early indication of a controller problem or a bad device.

We would say that if DBCCs indicated the table and database are clean, then there was a transient problem, which could have its roots in a disk-related failure beginning to show up, or a memory-related problem