Restrictions

There are several restrictions that apply to shrink operations on databases.

The shrink database operation may move pages saved as resume points for reorg subcommands (for example, forwarded row, reclaim space, and so on), reorg defragment, and similar utilities. The resume points for these utilities become invalid when these pages are moved. You can resume these utilities only at the beginning of the table or partition.
You cannot shrink master, model, archive, proxy, temporary, inmemory, or other databases with reduced durability.
You cannot include the off parameter with the alter database ... on, log on, and log off parameters.
You cannot run a shrink database operation simultaneously with dump database, dump transaction, or other alter database commands in a single database. The shrink database operation cannot run if these commands are in progress in the indicated database.
Shrink database operations fail if you start them while a dump transaction is already running. If you issue dump transaction while a shrink database operation is in progress, Adaptive Server terminates the shrink database operation to allow the dump transaction to proceed. You can restart the shrink operation by issuing the same command after the dump transaction completes.
You must run a shrink database operation in single user mode if you specify a page range to shrink only part of a device.
Shrink operations may create a hole when they remove space from the middle or end of a database. You can perform transaction log dumps after a shrink database operation that creates a hole in the data segment, although the transaction log may include a dependency on the data that occupied the hole when you load the transaction log.
Adaptive Server may issue error number 3193 and this message when you load the transaction log: “Database '<dbname>' does not have space allocated for data page <pageid> from the dump. Extend the database by <size> MB and retry the command”..

You may need to extend the database before some (and potentially, every) subsequent transaction log dump in the load sequence.

To avoid this error pro-actively, Sybase recommends that you perform a database dump immediately after you perform a shrink database operation.
To load a sequence of transaction log dumps that follow a shrink operation that did not include a subsequent database dump:
- Run load transaction
- If you cannot run load transaction because Adaptive Server issues error number 3193:
  1. Query sysusages to determine which holes require filling.
  2. Run alter database to fill the holes, extending the database to avoid error number 3193
  3. Run load transaction using the same transaction log dump
  Note: During these sequence of steps, the data holes you filled with the alter database command may be re-created by the load transaction command, and Adaptive Server may report error number 3193 error when you perform the next load transaction. If this occurs, repeat these steps for every load transaction you perform.
Shrink database operations do not update indexes marked as suspect on tables whose pages will be moved as part of the shrink operation. After the shrink operation completes, these indexes remain marked as suspect, and may include references to data pages that are part of a device hole. Running dbcc checktable on tables with suspect indexes may result in error number 806 (this is expected). Instead, drop and re-create these indexes, or repair them with reorg rebuild table or dbcc reindex(table_name, 16).