| How to recover the database when running only on buffer cache [message #684131] |
Thu, 08 April 2021 10:46  |
Andrey_R
Messages: 441
Registered: January 2012
Location: Israel
|
Senior Member |
|
|
Hi all,
I submitted
QL> alter database begin backup;
Database altered.
- RMAN backups are there from a day before but the archive logs were removed so I believe using it to restore today's data is not an option
- Data pump not working due to a already corrupted DBMS_REGISTRY and other SYS objects
The application system works fine, but queries on v$ get error
SQL> select * from v$database;
select * from v$database
*
ERROR at line 1:
ORA-00210: cannot open the specified control file
ORA-00202: control file: '/....../control01.ctl'
ORA-27041: unable to open file
Linux-x86_64 Error: 2: No such file or directory
Additional information: 3
SQL>
What other options can be used in such a situation, to save the database?
Thanks,
Andrey
|
|
|
|
|
|
| Re: How to recover the database when running only on buffer cache [message #684134 is a reply to message #684133] |
Thu, 08 April 2021 11:35  |
Andrey_R
Messages: 441
Registered: January 2012
Location: Israel
|
Senior Member |
|
|
Quote:Why did you remove those files?
User error ( not realizing which instance the deletion is done on ).
Quote:What type of backup was it (full, incremental 0, incremental 1?)
Full plus archivelogs
Quote:Why did you issue the 'alter database begin backup;'?
As a part of a datafile copy based repliaction scenario from one machine to another similar one
Quote:What version of Oracle? What OS?
12.2.0.1.0 , OEL 7.9
|
|
|
|
| Re: How to recover the database when running only on buffer cache [message #684137 is a reply to message #684134] |
Thu, 08 April 2021 13:03 |
John Watson
Messages: 8964
Registered: January 2010
Location: Global Village
|
Senior Member |
|
|
When you "deleted" those files, you didn't actually delete anything: all you did was remove the references to them from their directory, which unlinks them from their inode. As long as a process has the file open, it will still be able to access the file contents. Only when all processes have closed the file will the file actually be deleted. You will likely find that any new sessions cannot do anything, as they will not be able open any of the files.
What you should be able to do is use lsof to list the inodes of all open files. Then terminate the instance, and create a new temporary filenames pointing to them. Work out which files they are, and rename them appropriately. I don't believe in luck (because I am a scientist) but you are going to need some.
|
|
|
|
| Re: How to recover the database when running only on buffer cache [message #684139 is a reply to message #684137] |
Thu, 08 April 2021 13:36 |
Andrey_R
Messages: 441
Registered: January 2012
Location: Israel
|
Senior Member |
|
|
John Watson wrote on Thu, 08 April 2021 21:03When you "deleted" those files, you didn't actually delete anything: all you did was remove the references to them from their directory, which unlinks them from their inode. As long as a process has the file open, it will still be able to access the file contents. Only when all processes have closed the file will the file actually be deleted. You will likely find that any new sessions cannot do anything, as they will not be able open any of the files.
What you should be able to do is use lsof to list the inodes of all open files. Then terminate the instance, and create a new temporary filenames pointing to them. Work out which files they are, and rename them appropriately. I don't believe in luck (because I am a scientist) but you are going to need some.
Thank you very much.
the session was disconnected as I tried to issue "select scn_to_timestamp(current_scn) from v$databse".
Then hope was gone and a restore to the last rman backup was performed and database opened with resetlogs.
However, I did see that while the database was running based on "ALTER DATABASE BEGIN BACKUP.." - the files were gone - directories were empty.
When I tried to query some data from v$database from the session that issued the "begin backup" - it gave errors about sys..dbf file missing
Andrey
[Updated on: Thu, 08 April 2021 13:41] Report message to a moderator |
|
|
|
| Re: How to recover the database when running only on buffer cache [message #684143 is a reply to message #684137] |
Thu, 08 April 2021 15:15 |
Andrey_R
Messages: 441
Registered: January 2012
Location: Israel
|
Senior Member |
|
|
John Watson wrote on Thu, 08 April 2021 21:03When you "deleted" those files, you didn't actually delete anything: all you did was remove the references to them from their directory, which unlinks them from their inode. As long as a process has the file open, it will still be able to access the file contents. Only when all processes have closed the file will the file actually be deleted. You will likely find that any new sessions cannot do anything, as they will not be able open any of the files.
What you should be able to do is use lsof to list the inodes of all open files. Then terminate the instance, and create a new temporary filenames pointing to them. Work out which files they are, and rename them appropriately. I don't believe in luck (because I am a scientist) but you are going to need some.
I read and re-thought what you said and understand it a bit better.
Fascinating. I may create a testcase and try it at some point and post it.
Thanks again.
|
|
|
|
|
|
] 
Blog
Search
Help
Members
Register
Login
Home
