Nystel, December 15, 2002 - 11:11 pm UTC
When the database is not having any activity, will Oracle processes be still performing I/O's against the database? If yes then what process and what activity is it performing against the files?
Regards
December 16, 2002 - 7:38 am UTC
DBWR -- flushing blocks.
LGWR -- flushing redo buffers (every 3 seconds)
CKPT -- updating datafile headers
constrol files getting written to every 3 seconds at least...
yes, a database with no active sessions could be doing tons and tons of IO (writes).
what is there to flush?
GUY, December 16, 2002 - 1:17 pm UTC
tom you said
---------------
DBWR -- flushing blocks.
LGWR -- flushing redo buffers (every 3 seconds)
CKPT -- updating datafile headers
constrol files getting written to every 3 seconds at least...
yes, a database with no active sessions could be doing tons and tons of IO
(writes).
------------
but if there is no activity then
DBWR -- no blocks for flushing blocks.
LGWR -- no activity so no flushing redo buffers (every 3 seconds)
CKPT -- updating datafile headers -- update for what?
It would really help if you could explain with examples.
December 16, 2002 - 1:50 pm UTC
you do an update. you go away. you don't commit.
database has no active sessions now....
LGWR -- flushes redo from your update after a bit... sure, you haven't committed but your redo is flushed
DBWR -- times out, time for a checkpoint, finds dirty blocks from 1/2 hour ago and writes them out.
CKPT -- signalled by end of checkpoint to update datafile headers....
SMON wakes up and coalesces free space, generating log, dirty blocks and eventually causing a checkpoint. SMON does tons in the background -- all of the time.
Datafiles are constantly being written to, even when "nothing is happening"
Nystel, December 16, 2002 - 6:31 pm UTC
If there is no database activity then what information is being written into the control files every three seconds?
If the redo log buffers are not written to (since database is not active), then what is the reason to flush the redo log buffers? Does that mean when there is no db activity still we will see online redo log file switched at the same frequency (since redo log buffers gets flushed to online redo log files)?
Regards
December 17, 2002 - 7:45 am UTC
stuff -- just stuff. stuff we have no control over. look at your disk drive light and you'll see it like clockwork (it is clockwork stuff that is written).
SMON is doing stuff. SMON is generating redo. Your inactive database isn't INACTIVE, you have taken the premise that "if I'm not doing anything, it is not". well, that premise is wrong.
SMON is just one of them, SNPn's, intermedia, AQ, many other background processes out there with the potential to do stuff
redo logs will only switch as they fill -- smon may eventually fill one and switch it sure. will it be at the same rate as when you have tons of users? of course not, it'll be much slower.
suspend activity
Baqir Hussain, January 21, 2003 - 7:10 pm UTC
Tom,
When the database was put into suspend mode
alter system suspend;
then any kind of activities being happened to redo logs and control files??
can we achive a consistent redo logs and control file during the following:
1. Alter tablespace user begin backup;
2. Alter system suspend;
3. copy data , redo and control files
4. alter system resume;
5. alter tablespace user end backup;
January 21, 2003 - 7:34 pm UTC
You will NOT achieve consistent backups, they will be HOT FUZZY backups.
The only way to get "consistent" (and here consistent is not a goal, not a "good" thing) is a cold backup after a normal or immediate shutdown.
Those are normal (good, desirable) fuzzy hot backups. I would NOT use suspend/resume for that purpose. only for the purpose above.
A reader, September 16, 2003 - 5:21 pm UTC
How is db consistency achieved with tripple mirrors silvering/resilvering
Anil, September 02, 2004 - 4:23 am UTC
Hey Tom,
1. Can you please explain how three way mirrors work, in essence how is the database consistency achieved?
Also:
2. On idle db: We had a simple test, we had a dummy database rnning for three days, doing nothing, it was just there, no queries, no dml nothing whatsoever. we copied the dbf's online got rid of those files, restored them with the earlier "backup" and opened the db, and it opens. Was curios, if something being written to, I thought our backup shouldn't have worked at all! Please help!
Really appreciate your inputs.
September 02, 2004 - 8:54 am UTC
1) to us, to the database, raid mirroring is totally 100% transparent, we don't "see it"
You have to ask the hardware vendors, the raid makers how they do it :) to us, it is just a disk...
unless you are talking about ASM in 10g, then we do it. We do it by simply maintaining 1 or 2 copies (mirror or triple mirror) of extents on different failure groups.
2) you can get "lucky", you should never count on getting "lucky".
you got "lucky" is all.
PMON
Shahid I Siddique, September 24, 2004 - 1:40 am UTC
My question is regarding backgroung processes.
PMON releases the lock and resources held by user and is managed by Oracle Server, right!!
Is there any way by which a DBA can "control" the PMON?
Thanks in advance
September 24, 2004 - 10:48 am UTC
in which circumstances?
no, pmon does not "releases the lock and resources held by user", the user does that normally.
Alter system suspend/resume usefull when using snapcloning?
PaulS, March 04, 2005 - 6:25 am UTC
I wonder if suspend/resume is a method to use for making full database backup's using snapshot/snapclone?
What is an acceptable period of time between the suspend and the resume command? Making a snapshot wil take some 2-3 minutes.
March 04, 2005 - 8:22 am UTC
depends on the vendor, otn has papers on this, you would want to work with you vendor (hardware) on this as well.
alter system suspend
abz, October 15, 2007 - 10:36 am UTC
Version 10g Release 2 and 11g Releae 1.
Please confirm this statement.
The database is up and running, I issued
ALTER SYSTEM SUSPEND, note that I did not made any tablespace in BEGIN BACKUP mode at all (not before suspend command and not after suspend command).
are the all the following files be freezed:
1- All controlfiles
2- All datafiles
3- All online redolog files.
Can I copy all the above three type of files in this state to another machine and startup the database at that new machine.
October 15, 2007 - 12:35 pm UTC
well, if you are going to take a copy at that point, why not just shutdown abort.
the suspend is for breaking BCV's (business continuity volumes). It is not really intended for copying the entire database while suspended - the database will appear totally broken to everyone - it would be a 'bad idea'.
we are using flash copy
abz, October 15, 2007 - 3:10 pm UTC
If shutdown was possible for us, I didnt need to ask this question, but we dont want to shutdown the database for some reason.
I understand its for BCV, but I just want to know that,
are all control files , datafiles and online redolog files
on disk are freezed after suspend? Please give answer of this specifically.
By the way we are using IBM flash copy to copy the files.
May be you will say to ALTER DATABASE BEGIN BACKUP, and then copy files and archived log files and apply the archived logs on target, but we also dont want to apply archived logs for some reason.
SUSPEND I/O
abz, October 17, 2007 - 10:51 am UTC
Why is file INPUT and OUTPUT both are suspended in SUSPEND?
Why not only suspending file output is sufficient for the purpose SUSPEND serves.