From Timothy.Clarkson@shell.com Thu, 05 Jun 2003 01:24:32 -0700 From: "Clarkson, Timothy T SEOP-OEIRH/1" Date: Thu, 05 Jun 2003 01:24:32 -0700 Subject: RE: Good technical documents/references on tuning restores Message-ID: MIME-Version: 1.0 Content-Type: text/plain Title: RE: Good technical documents/references on tuning restores I had this problem yesterday.  not sure if what I did is correct but it seemed to work for me.   1) Create another tablespace (TEST) 2) move the objects from the tablespace that has the missing files into TEST tablespace     select DISTINCT TABLESPACE_NAME from dba_extents where file_id in                 (select File# from v$datafile where status = 'RECOVER')   you might have some problems with items that resided on the missing datafile, in my case it was only indexes so I recreated them from the schema on the new tablespace.     select DISTINCT SEGMENT_NAME, SEGMENT_TYPE, TABLESPACE_NAME from dba_extents where file_id in                 (select File# from v$datafile where status = 'RECOVER') 3) drop the tablespace with missing datafiles 4) move back the objects to the original tablespace   This cleared up the mess in the v$datafile.  I did this because I was under the impression that you cannot delete a datafile once it has been assigned to tablespace.     Have not got around to missing the missing file in the TEMP space yet, as it is a test database will do that later.   Hope this helps, I am sure someone will post a better answer soon :-)   Tim Clarkson -----Original Message---- From: Henry Poras [mailto:[EMAIL PROTECTED]Sent: Wednesday, June 04, 2003 3:10 PMTo: Multiple recipients of list ORACLE-LSubject: RE: Good technical documents/references on tuning restores Paula,   IIRC 'MISSING' means it is in the data dictionary, but not the control file. I've gotten that before, but haven't played that game in a while. You might be able to 'DROP' them, but I'm not sure. Check that first on a mini-test system.   Henry   -----Original Message-----From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]On Behalf Of [EMAIL PROTECTED]Sent: Tuesday, June 03, 2003 11:05 PMTo: Multiple recipients of list ORACLE-LSubject: RE: Good technical documents/references on tuning restores Okay,   Database open and I wanted to not restore and not reference certain index and materia. view datafiles.  Now the datafiles show up with weird MISSING...... names.  How do I clean this up?   Thanks, paula -----Original Message-----From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]Sent: Tuesday, June 03, 2003 10:20 PMTo: Multiple recipients of list ORACLE-LSubject: RE: Good technical documents/references on tuning restores Strange: looking at v$backup_async_io found some datafiles where         active bytes per sec            long-waits        2154608                          297            7489829                                297  How can I start finding the disparity?  -----Original Message----- From: Stankus, Paula G Sent: Tuesday, June 03, 2003 9:05 PM To: Stankus, Paula G; '[EMAIL PROTECTED]' Subject: RE: Good technical documents/references on tuning restores Did I say that throughout this restore/recovery process my tapes being mounted and unmounted while other database backups and cloning was occurring.  Big duh - this has got to be slowing down the process.  Any ideas on how to handle this?  Perhaps more than one tape subsystem would be better or while doing restores other backup utility is halted - hmmm - but that risks those other systems?  How can I sell that?  Hmmmm.  Maybe I will first backup to disk.  Yep, one backup to disk.  Yep..... -----Original Message----- From: Stankus, Paula G Sent: Tuesday, June 03, 2003 9:00 PM To: Stankus, Paula G; '[EMAIL PROTECTED]' Subject: RE: Good technical documents/references on tuning restores See nwadmin shows parellism = 32.  I am not the tape person - that is my sys admin.  I don't believe they actually have 32 tapes involved only 4.  Can this impact performance by setting this too high?  I also notice one backup server for networker with nwadmin messages like "clone sets being created...", "other databases being backed up".  What is the usual policy on concurrently running backups, cloning at the same time recovery is taking place.  It appears to show a lot of media waiting events:  waiting for dlt7000 tape ... of the tapes that I specifically need for this restore.  Did I say:  I love my admin...I love my admin....I love my admin... -----Original Message----- From: Stankus, Paula G Sent: Tuesday, June 03, 2003 8:53 PM To: Stankus, Paula G; '[EMAIL PROTECTED]' Subject: RE: Good technical documents/references on tuning restores Okay, almost 4 minutes for a datafile that was 78168 blocks - how do I know if this is reasonable? Also, seems to write these files out (restore sychron.) why can't it restore different datafiles in parallel?  - stupid question huh? -----Original Message----- From: Stankus, Paula G Sent: Tuesday, June 03, 2003 8:46 PM To: Stankus, Paula G; '[EMAIL PROTECTED]' Subject: RE: Good technical documents/references on tuning restores SELECT SID, SERIAL#, CONTEXT,                             ROUND(SOFAR/TOTALWORK*100,2) "% complete",              SUBSTR(TO_CHAR(SYSDATE,'HH24:MI:SS'),1,15) "Time now"    FROM V$SESSION_LONGOPS                                   WHERE OPNAME like '%restore%';                           .                                                        showed all 100% complete but msglog from RMAN shows it is truly still running.  -----Original Message----- From: Stankus, Paula G Sent: Tuesday, June 03, 2003 8:40 PM To: Stankus, Paula G; '[EMAIL PROTECTED]' Subject: RE: Good technical documents/references on tuning restores Also found on monitoring performance of RMAN jobs: Note:144640.1 on Metalink -----Original Message----- From: Stankus, Paula G Sent: Tuesday, June 03, 2003 8:36 PM To: Stankus, Paula G; '[EMAIL PROTECTED]' Subject: RE: Good technical documents/references on tuning restores Found this white paper: http://otn.oracle.com/deploy/availability/pdf/rman_performance_wp.pdf Anything better? -----Original Message----- From: Stankus, Paula G Sent: Tuesday, June 03, 2003 8:33 PM To: Stankus, Paula G; '[EMAIL PROTECTED]' Subject: RE: Good technical documents/references on tuning restores This is what I have set on my target database: ----------------------------------- ------- ------------------- backup_tape_io_slaves               boolean FALSE              tape_asynch_io                      boolean TRUE               Version 8.1.7.4 database and RMAN catalog - 32 bit Networker MML Using RAID 1+0 Solaris 2.8 -----Original Message----- From: Stankus, Paula G Sent: Tuesday, June 03, 2003 8:28 PM To: Stankus, Paula G; '[EMAIL PROTECTED]' Subject: RE: Good technical documents/references on tuning restores Okay - from my reading you don't need to have multiple tape io slaves if you are using asynch. I/O.  Again, best document for perf. tuning database restores using RMAN would make mucho difference.  Read old note about someone doing an analyze on the RMAN catalog tables to improve performance of restore.  I think it has something to do with how quickly it finds the file on tape and writes to disk.  -----Original Message----- From: Stankus, Paula G Sent: Tuesday, June 03, 2003 8:25 PM To: '[EMAIL PROTECTED]' Subject: RE: Good technical documents/references on tuning restores Seems to be taking awfully long to read files from tape and write to disk.  I allocate multiple tape channels like I do for the backup which only takes about 45 minutes.  Does not seem to be spawning multiple sessions.  Do I need to change parameters on my init.ora file to use multiple tape io slaves to see this. Anyway, would like notes/docs., references if you all have some. Thanks, Paula