Is there a way to simulate Oracle VSS writer behavior in pre/post snapshot scripts?

User_20ACP

We use VSS to backup our Windows servers. When these server contain an Oracle database, we used to use the Oracle VSS writer to ensure a consistent database. Unfortunately with Oracle 12c and the introduction of pluggable databases, the Oracle VSS writer no longer functions correctly. It contains a bug that causes a failure when it attempts to put a pluggable database into backup mode; see https://community.oracle.com/tech/developers/discussion/comment/15467069

Since it's unclear when/if this issue is going to be fixed, I'm looking for ways to simulate what the Oracle VSS writer does in scripts that run "pre/post" when the VSS snapshot is taken. In the "pre" VSS snapshot script, I use RMAN to backup the control file, run "alter system archive log current" to get a new archive redo log, and then put the database in backup mode with "alter database begin backup". After the snapshot is taken, the "post" script runs "alter database end backup". The recover process uses "recover database using backup controlfile until cancel", but fails with an ORA-01194 file needs more recovery error. This error makes sense because the "alter system archive log current" really needs to be run after the database is taken out of backup mode, but if done there the archive redo log created would not be included in my VSS snapshot (and not available for recovery).

I'm not really sure there's a way around this... However, I'm curious if anyone knows what the Oracle VSS writer does internally since if it didn't have the bug above (as is the case in older versions of Oracle that don't support pluggable database), it would allow for proper recovery?

Note that I'm able to perform an "incremental forever" backup in the "pre" snapshot script and use those files to restore the database and do point-in-time recovery for the archived log files, but I'm trying to avoid having to consume storage/restore time required for a backup of the database files.

Any help is greatly appreciated.

Christian.Shay -Oracle

Hi,

I don't have a solution for you now, but just wanted to say that the VSS Service not starting bug will be fixed in the next release of XE for Windows.

User_20ACP

Thank you for the response. Is there a link to this bug so I verify that the issue we are seeing is the same as the "VSS Server not starting" bug?

Christian.Shay -Oracle

https://community.oracle.com/tech/developers/discussion/comment/16784961#Comment_16784961

Hi, no I don't have a page describing this bug, but it is very simple: XE 18c on Windows includes a Windows service called "OracleVSSWriter". This service is responsible for quiescing the database so that Volume Shadow Copy can take a consistent snapshot. Without that service running, you can't use VSS properly.

Unfortunately in XE 18c, this service does not start. We have fixed this in the next release of XE on Windows.

User_20ACP

Christian,

Thanks for the reply. We do not see this problem using 12c or 19c (I haven't tried 18c). For 12c and 19c, the OracleVSSWriter starts but when a VSS snapshot is taken, the OracleVSSWriter crashes resulting in an unrecoverable set of database files. The following thread https://community.oracle.com/tech/developers/discussion/comment/15467069 discusses another issue where the VSS Writer fails attempting to put a pluggable database in backup mode (which I suspect we'd encounter if the OracleVSSWriter didn't crash). I'm curious if these issues are also addressed? Does Oracle publish a list of outstanding bugs in Oracle database components?

Steve