That statement is often written by Jonathan Lewis and today I was reminded on how true it is.
I had a discussion today with two of my colleagues who wanted to increase the number of arch processes on a dataguard system. As reason they pointed to metalink note 468817.1 – “RFS: possible network disconnect while taking rman backup on primary site”, which makes the following statement:
“In a Data Guard Configuration, during Scheduled RMAN Backup no Redo is transported to the Standby Server as the ARCH Process is blocked (as expected ie. RMAN would utilize 1 ARCn Process and the other ARCn for local Archiving ) which means the Standby stays out of sync (assuming max_arch_processes=2) until the ArchiveLog is manually copied across and registered to the Standby Database after which the Standby Database resumes applying the ArchiveLogs.”
The first thing that drew my attention was the part about rman utilizing an ARCn process.
While the ARCn processes are indeed responsible for archiving the online redo log files, it is the sessions own server process that does this when issuing an “alter system archive log current” command. When rman forces a log file to be archived, the same thing happens. This can easily be verified by looking to the “creator” column in the v$archived_log view. A strace of the rman server process would also prove that it is this process which reads the archived redo logs and streams them to the rman client to be written in a backup piece.
1 – 0 for me
The second thing I noticed was that, according to the note, the standby would remain out of sync until the archivelog was manually copied across and registered to the standby db.
Even if the rman process was using an ARCn process, leaving no processes to copy the archivelog over (with max_arch_processes=2), the standby db would normally be able to pickup the synchronization again after rman would have released the ARCn process again.
2 – 0 for me
and end of discussion.
So, “just because its printed, doesn’t mean its true”, would also apply on metalink notes.
Luckily there is a feedback link at the end of the note, so I hope it will soon be removed or modified.
And yes, I can get a little bit competitive in discussions.
How did you guess?
