SMON instance recovery

Similar Questions

• How instance recovery and restart the database after SHUTDOWN ABORT?

  How instance recovery and restart the database after the database is closed by using the command "SHUTDOWN ABORT"?

  Recovery if instance Oracle background SMON instance failure process using or stop quitting or database starting strength. After shutdown abort you just restart the startup database, oracle will take care for instance recovery you need'nt to intervene for example recovery.

• Can someone help me understand the process Instance recovery

  First thanks for all of you who responded to my post, any help is really appreciated.
  
  I would like to more deeply understand the Instance recovery process then the Administration Guide of Oracle Docs can tell me.
  
  What I have so far of Instance recovery is:
  
  1 oracle determines which files are not synchronized.
  2. all data from redo log files is applied to restore log buffers by progression. Buffers do not participate on the database of normal transactions because they are marked to pick it up.
  All these data are applied to the data files (commited and contains transactions).
  This step is called roll forward.
  3 cancellation are used to restore transaction no. After this step the datafiles not aura no data contains on them. At this point, the instance is available for users, but the blocks used are not.
  4. all files are now in sync.
  
  My questions are:
  
  -Step 1, what process determines which files are not sync? Make the bed of Smon the controlfile and data file headers and then make a list which is not sync?
  
  -L' step 2, what is the process that reads the information in the redo log files and complete the recovery log buffer? What is the CKPT? And how he is warned that he might start to read the files after completing step 1? There is a semaphore that is defined for it?
  
  -Step 2, which process applies information engaged and contains buffer redo log for the data files? When it starts to apply this information of the instance is already partially available to users?
  
  -Step 3, what is the process that reads up to cancel to restore transactions not? This process must be done once step 2 is complete as could be blocks of the undo tablespace on buffers of the right-wing newspaper do it again? And when this step 3 starts, the process is triggered by an another semaphore?
  
  -When step 3 is finished, what process changes the State of the database of partially open to open, or that's just what we call it when the database has always revenge marked for recovery?
  
  Thanks again in advance for the help, there are reasons not practice known this information, I'm just a curious person who could not find this information on the google...
  
  
  Ricardo Rodriguez

  Ricardo,
  Well, I'll take a chance to respond.
  (1) SMON in the start-up phase, in the strict sense in the editing phase, compares the SNA stored in the control file with the data file headers. This sCN is called checkpoint SCN. It is always higher among the controlfile so its part of the same database and is not set up to an old backup. By this correspondence, Oralce comes to know that there are some report files with their SNA control point is only not with the controlfile and consequently, instance recovery is reported. Here is the list of files is prepared.
  (2) the same process SMON only would trigger playback of the redo log files and launch the Rollforward process. Roll forward is made from the last address bytes redo controlled at the end of the restore log forward. This means that the last block, blocks have not reported in the data file are simply read and are dumped on the data files. As a result, we get a situation that the data file will end up with a lot of data committed but also committed that redo is containing both.
  CKPT process will not come in the picture I think. Where did you read what CKPT would do all this?
  (3) SMON governs all this thing is to say the application of the data for the data files. No, in this phase, we are still in the phase of mounting that this recovery is called crash recovery. Oracle wouldn't let users to use the database at this point. Data is still not available for them.
  (4) do not bring semaphores here. They are already in use when you allocated the memory of your instnace. Its work process level. After the mount phase is completed, oracle would open up the db. At this stage, "the transaction recovery" aka rear roller phase. That would be launched to relax all the operations that are not validated.
  Now to do this, undo will be used. Oracle launches this phase into two or more strictly speaking 3 ways. One, SMON would just check how many blocks will be there for the recovery. If they are larger, it would begin the rear roller with parallel slaves. This is called "we demand" rollback. It is initiated by the user. So if you were using the EMP table data and the server crashed between, now the next time you start, if you try to access the same blocks once again, they are set aside on the basis of priority.
  The third one will rollback transaction intra. It's basically splitting the work among the slaves so that if a slave is sitting free, it can share the work and can move forward to make things more quickly restored.
  Please note that all this is happening at the open stage. This means then that this is the case, oracle is available for users. And second, there is little change in this process (an integer) 9i go. This is called two-pass recovery which has been made to make this process a little faster. In this regard, SMON ignores these blocks that are controlled by DBWR during execution of the phase of rollforward.
  Hope this makes some sense and helps. Let me know if there is something wrong or unclear.
  HTH
  Aman...

• instance recovery

  Hello
  
  With regard to the recovery of instance in an envionment of the cars,
  
  Parallel recovery can still operate even if the recovery_parallelism the value 0 or 1 initialization parameter?

  Ora_83 wrote:
  Thanks Satish,
  This means that parallel instance recovery does not work if I put the recovery_parallelism initialization parameter 0 or 1?

  I have no reason to think otherwise.

• Instance recovery fast and redolog file size

  Hello
  
  Could you please explain to me how redolog file size allows to retrieve the instance more quickly?
  
  Thank you
  KSG

  Soon, I'll try to say.

  The answer lies in the relationship between the number of buffers Sales necessary for recovery. Buffers dirty necessary for recovery is limited by the checkpoint process when DBWR is ping some of them writing in the data file. So with a log switch, you'll hit a control point. So if the size of the log files to be smaller, frequency control would be great, which makes written buffers more aggressively in the fikles of data and limit the duration of the reocvery instance. Her biger make you, it would take more than the checkpoint occurs and so in the case of instance recovery, it would bemore time required. That said, using a log file of small size would also lead to "cehckpoint inicomplete" error as well, because it can happen that DBWR will not be able to match with his own writing speed and the speed of the current generation gate of event.

  HTH
  Aman...

• Ora-00265 instance recovery required to CCR after immediate cessation

  Hello world

  I have this problem in a database of production, (version 11g 2, 11.2.0.3) using Real Application Cluster, I disabled archivelog to a window for a mass of large tables update, now I'm trying to reactivate the archivelog mode and I couldn't, because it shows the error ORA-00265, I find a lot of sites and Oracle support and the explanation is always the same "you shutdown abort or force start the base of. data start again and stop with immediate or normal option"but the thing is that I use srvctl with the immediate option, I did a few reboots of the database and I read in the newspaper of the alerts that indeed there at the start of the database runs a recovery of Instance.

  Why does this happen? I stop the two nodes at once and one by one, began with sqlplus and I didn't really know what else to do, I hope you can help me.

  Oracle@sur01 # srvctl stop asaa d-i o asaa1 immediate instance

  Oracle@sur01 #. oraenv

  ORACLE_SID = [oracle]? ASAA

  The Oracle base has been set on/u01/app/oracle

  Oracle@sur01 # export ORACLE_SID = asaa1

  Oracle@sur01 # srvctl start database d asaa o mount

  Oracle@sur01 # sqlplus / as sysdba

  SQL * more: Production of the 11.2.0.4.0 version sat Jan 24 15:48:16 2015

  Copyright (c) 1982, 2013, Oracle.  All rights reserved.

  Connected to:

  Oracle Database 11 g Release 11.2.0.4.0 - 64 bit Production

  With Real Application Clusters and Automatic Storage Management options

  Instance Estado Modo Lectura Activa Actividad Blo Inicio Archive?     EstArch

  ---------------- ------------ -------------------- ----------------- --------- --- -------------- ------------ -------

  MOUNTED asaa1 MOUNTED NORMAL ASSET No. 24/01/15 15:47 NOARCHIVELOG ORDER

  SYS@asaa1 > changing archivelog database;

  ALTER database archivelog

  *

  ERROR on line 1:

  ORA-00265: recovery required instance, cannot define in ARCHIVELOG mode

  Thanks in advance...

  Giancarlo Giammaria.

  Hello world

  To share knowledge, I write what I had to do to solve the problem:

  It seems that the FCAC does not correctly close instances (another question to consider), causing the datafile headers remain incompatible with a different number of SCN.

  In the next query, you can see as the checkpoint_change # in v$ database is different from the checkpoint_change # in v$ datafile (this question was scheduled to be executed in mounted State)

  SYS @ > select dbid, name, created, resetlogs_change #, resetlogs_time, open_mode, log_mode, checkpoint_change #,.

  2 controlfile_type, controlfile_change #, controlfile_time

  3 v $ database;

  NAME DBID CREATED RESETLOGS_CHANGE # RESETLOGS_TIME OPEN_MODE LOG_MODE CHECKPOINT_CHANGE # CONTROL

  ------------------------------ ---------- -------------------- ------------------------------ -------------------- -------------------- ------------ ------------------------------ -------

  CONTROLFILE_CHANGE # CONTROLFILE_TIME

  ------------------------------ --------------------

  2103866247 XXX 11 OCTOBER 2014 12:00:07 1 OCTOBER 11, 2014 12:00:07 MOUNTED NOARCHIVELOG 332020165 CURRENT

  332058714 26 JANUARY 2015 22:37:34

  1 selected line.

  SYS @ > select the folder, name, State, checkpoint_change No. # from v$ datafile;

  FILE NAME STATUS CHECKPOINT_CHANGE NO. #.

  ------------------------------ ---------- ------- ------------------------------

  332020165 SYSTEM 1 + DG_DATA/a

  2 + DG_DATA/a 332020165 online

  3 + DG_DATA/a 332020165 online

  Then, the process that I did to fix that and enable archivelog:

  srvctl stop database d immediate o xxx

  srvctl start database d o mount xxx

  recover the database;

  srvctl start database-d xxx (apparently the bd still needs a clean shutdown to reactivate the archives, otherwise the ORA-00265 continues to show, so I had to start and stop)

  srvctl stop database d immediate o xxx

  srvctl start database d o mount xxx

  ALTER database archivelog (at this point, it worked and the archiving log enabled, after this, stop and open again...)

  Hope this helps someone else!

• Instance recovery is possible without regrouping

  Sorry if this is in the wrong place, but it is a related issue of RAC.  I usually only ever work Linux or Solaris.  Someone asked me to set up a database for them under Windows and is unwilling to use clusters which I am suspicious of as well, Windows can have its problems. Recover a database sure thing, no problem.  An instance goes down in a cluster and then no problem. However, if you have one instance and which gets corrupted, what is the fastest method to recover the incompatible database? In this scenario, it is just the Forum I wonder about for example a corrupted etc. registry where the instance just does not start but the spfile, file control, data files, recovery online newspapers are archived files log files are fine.

  Sorry if it's a question of OS for many it's just that all literature I've ever read is on the covers of high availability or the protection of the database.  Can't seem to find anything about the fastest way to get a fixed corrupt single instance and addressing the existing database again.

  Kind regards

  Dave.

  Is this is a scenario where the registry or perhaps the window system is corrupt these types of emissions should be solved with a certain type of backup at the BONE level rather than a product of Oracle. Once this type of corruption is resolved, then you should be able to restart the database if the data files, the redolog and the controlfiles are very good.

  You may also be waiting on a separate server where you could then failover and point the application to the database pending.

  Best practices on windows is to have separate physical disks for operating system files, Oracle software and database.

  Example of

  C:\ - OS

  E:\ - software Oracle<- local="" or="">

  F:\ - database Oracle files<- local="" or="">

• A 2 DB instance recovery catalog

  Hello
  
  I have a up and race catalog rman (catdb) recovery with a recorded case (orcl) in it. How to add another instance (devdb) in my recovery catalog (catdb)?
  
  Thank you

  You can register a number n of database in the recovery catalog database.

  connect to the database of the registry database and recovery catalog;

  for example:

  Catalog RMAN rman/rman@catalog_database_name target = sys/password@target_db_name =

  RMAN > register database;

  To connect to the catalogue of recovery and the query database;

  SQL > select * from db;
  for example:
  Connect on the demo1 recovery catalog database and as a catalog of recovery user sai;

  SQL> conn sai/sai;
  Connected.
  
  SQL> select * from db;
  
      DB_KEY      DB_ID CURR_DBINC_KEY
  
  ---------- ---------- --------------
  
           1 3710360247              2
  
         141 2484479252            14
  

  check out the link:
  http://oracleinstance.blogspot.com/2009/12/how-to-create-register-add-unregister.html

  Published by: rajeysh on August 16, 2010 10:16

• Oracle concepts: Clarity needed on the recovery Instance

  One of my colleague was discussing the recovery instance with me.
  We've discussed on the position of the control point, the phases of recovery of body (front bearing rear bearing), and then he asked me what is the point of rolling toward the before (application of redo logs) and then roll back (undo blocks application), why not apply only committed transactions?
  
  I told him that this happens because:
  (1) roll forward to recreate the scenario just before what Instance has crashed
  (2) roll front/roll back is the fastest way to recover
  
  I don't know where I got this learning, maybe some blog or article (probably not), but is it correct?
  
  Kind regards
  Ankit Rouault

  NVO wrote:

  The 2nd point is completely wrong!

  Let me clarify the frame of the 2nd point, it was mentioned in the context of the application only transactions posted during recovery. As all vectors of change go in the redo log files, there is a mechanism to identify only committed transactions, it would have taken more time (its really hypothetical because there is no mechanism exists, right?) by this mechanism to identify transactions of dedicated (from newspapers of recovery) that just roll forward and back.

  There is no context at the moment, that's wrong , plain and simple!

  There is a mechanism to identify committed transactions and called commit . But when you are recovering, you don't recover only transactions posted but you re-create the scene before the crash, which means, you would be required to apply the two vectors of change / committed and dedicated work. By your logic, if Oracle would start looking for committed transactions, it would be a lot of questions. One, time needed would be higher. Two, there may be an instance validation has been issued but before that the full validation message could reach the user, the Forum got arrested while is those blocks that can be today in the data of the file and have obtained a status of active transaction and are not sure whether or not they are committed. And finally, the LGWR writes way is just to continue to push for change vectors in the flow to restore order of the CNS. So to do a restore, you must reapply all these data that have been changed in this same order of SNA.

  Instance recovery means that you do not encounter a physical file is overwritten or deleted but there is a loss of memory and some tampons were waived to be written in the data files and now, data files, control file and the log files are not synchronized.

  The second point, where you say that it is the fastest, is wrong, because even for the resumption of the proceeding, it can take a lot of time and depends on the amount of work has been left before the instance has been broken. Roll forward is the way to put data files synchronized with the redo log files and control files. So if the data files were about control number 100 and control files and redo files log recording the number like 120, the missing gap of 20 numbers would be reinstated in the data files because only then, the database would be open. Now, your friend doubt that why not only committed transaction because you never say that a transaction is made only when there is a validation. Whenever any DML is encountered, it will cause a transaction. And you don't want that plenty of work to be recreated to ensure that, if there was a commit issued before the instance has been broken, the next time, Oracle would ensure that the work is there and if not, he'd be back. And above all, in redo log files all agents of change, not only for committed transactions.

  That's fine, even to understand here.

  Bring your own thoughts now that how you understood this explanation, until I started to get into more technical terms.

  I would like to know your opinion on 1 point. Assume that something went wrong (in terms of transactions) because the instance has crashed, if he rolls forward (as it happens automatically) on the next boot, it will crash again?

  He would return for the (next) start the instance. If all you're doing are plain DML, they don't plant the instance as long as you have not met any bug in the case that, next time also the instance may crash.

  Aman...

• How oracle will come to know weather it is recovery of Instance or Media Recovery

  How to know initially weather oracle's recovery of Instance or Media Recovery
  
  Oracle will make recovery Instance automatically, and I know that the process
  
  But how it identifies the weather it's Instance or Media Recovery recovery

  There is no recovery Instance during the START-UP phases and the MOUNT. Instance recovery is the opening time.

  Check the log of alerts of your database where you did a cover of the Instance (for example after a kill or reboot or shutdown abort) and see where appear messages of recovery of instance- after the ALTER DATABASE MOUNT is completed.

  You can STARTUP NOMOUNT with no present controlfiles.
  You can RIDE with no current data files.
  It is OPEN only which checks the data files.

  Controlfiles are never "found". They are either manually, restored or created by the DBA.

  Note: Instance recovery is called "crash recovery" in the alert.log file

  Hemant K Collette
  http://hemantoracledba.blogspot.com

  Published by: Hemant grapple on August 26, 2009 16:45

• Instance - Redo and Undo recovery question

  I'm a little confused about how to undo and redo used to restore, please help me to clarify this point. If redo records changes (validations and database changes to cancel the revenge), don't do it again and undo all come from the log to roll forward to instance recovery? What is the source of the undo data, do it again save or cancel the segments that have been rebuilt using recovery logs?
  
  Matt

  In simple terms:
  Source of redo - redo logs, source of cancellation - undo tablespace. Oracle applies again to the data files, including cancellations of TBS, first, then it restores the unvalidated data using TBS of cancellation.

• Oracle Instance crashes during the allocation of channels using the Oracle database backup Service

  I am running Oracle Version 11.2.0.3 on a 64 bit Windows Server 2008 SP 2 OS.  Today when I tried to allocate a channel for 'SBT_TAPE' my instance has crashed.  Everyone knows this?

  Set ORACLE_HOME=D:\app\oracle\product\11.2.0\dbhome_1\

  Set ORACLE_SID = CNFCRAC1

  Set SBT_LIBRARY=D:\app\oracle\product\11.2.0\dbhome_1\BIN\oraopc.dll

  Set OPC_PFILE=D:\app\oracle\product\11.2.0\dbhome_1\database\opccnfcrac1.ora

  RMAN > Connect target

  RMAN > run {}

  allocate channel t1 'SBT_TAPE' device type

  PARMS ' SBT_LIBRARY=D:\app\oracle\product\11.2.0\dbhome_1\BIN\oraopc.dll, ENV=OPC_PFILE=D:\app\oracle\product\11.2.0\dbhome_1\database\opccnfcrac1.ora)';

  }

  Windows Event Viewer:

  The failing application name: ORACLE. EXE, version 11.2.0.3

  Module failed name: oraopc.dll, version 12.1.0.0

  Exception code: 0xc0000005

  Offset: 0x00000000000249af

  ID of the process failed: 0 x 1510

  Failed application start tiem: 0x01d0eb098ecd2d2f

  The failing application path: d:\app\oracle\product\11.2.0\dbhome_1\bin\ORACLE. EXE

  Flaws Path E-mail: D:\app\oracle\product\11.2.0\dbhome_1\BIN\oraopc.dll

  I made sure that the PATH environment variable has the following entry.  D:\app\oracle\product\11.2.0\dbhome_1\BIN

  RAC1 Table of contents alerts journal:

  Fri Sep 11 09:04:23 2015

  Thread 1 Advanced to record the sequence 11277 (switch LGWR)

  Currently Journal # 3 seq # 11277 mem # 0: +REDO/cnfcrac/onlinelog/group_3.258.824214591

  Fri Sep 11 09:04:31 2015

  Archived journal 23067 extra for each sequence 1 11276 0x82a6e611 dest ID thread entry 1:

  Fri Sep 11 09:06:20 2015

  Reconfiguration began (old Inc. 16, new Inc. 18)

  List of instances:

  1 (myinst: 1)

  Fri Sep 11 09:06:44 2015

  Global Resource Directory frozen

  * death detected instance - invalid domain 0 = TRUE

  Restored communication channels

  Diffuse master hash resource bitmaps

  Blocks of non-local process cleaned

  Fri Sep 11 09:06:44 2015

  LMS 0: GCS shadows cancelled 0 1 closed, 0 Xw survived

  Fri Sep 11 09:06:44 2015

  Shadows GCS LMS 1: 1 cancelled, 1 closed, 0 Xw survived

  Set of master node info

  Subject to all requests for remote enqueue

  DWN-cvts replayed, questionable VALBLKs

  All enqueues grant awarded

  Post SMON start 1 IR pass

  Fri Sep 11 09:06:46 2015

  Instance recovery: looking for dead threads

  Beginning of thread 1 instance recovery

  Subject to all requests for remote-cache GCS

  Post SMON start 1 IR pass

  Fix write in SCM resources

  Contents of the journal of alerts 2 RAC

  Fri Sep 11 08:57:47 2015

  Archived journal 23064 extra for each sequence 2 11793 0x82a6e611 dest ID thread entry 1:

  Fri Sep 11 09:06:42 2015

  Starting ORACLE instance (normal)

  LICENSE_MAX_SESSION = 0

  LICENSE_SESSIONS_WARNING = 0

  Monitoring of normal startup parameters... I looked in the CRS, ORAAGENT, OCTSSD, and CRSD marbles, but I can't find anything that stands out as to what caused the end of abnomal, another of the communication between the two bodies has not responded within the specified time.

  -Jen

  Hi Jen,

  There is a known issue with Windows module. We are working on fixing it and the new module should be available soon. I'll update as soon as we release the fixed module.

  Sridhar

• Restart weblogic cluster instances of command line

  Hi team,

  We have an obligation to cluster weblogic instance recovery every week.

  Currently, it is a manual activity by following - to connect to the console of the server admin-> go to Clusters-> choose Cluster-> restart all the cases one by one.

  We have 8 instances and 4 application servers [1:2 ratio], each application server succeeded the server node, which is responsible for 2 instances

  As far as I know, we can start/deploy applications through deployment using WLST/weblogic. The Deployment Manager

  Please can you suggest a UNIX utility by which we can restart the weblogic directly from unix/linux server cluster instances.

  Thanks in advance.

  Hello

  You can use WLST to the power via the script Manager node as you do it manually it forms the console. You can start the entire cluster at the same time:

  Start ('CLUSTER1', 'Cluster');

  or server by server:

  Start ('C1_MAN_01', 'Server');

  Start ('C1_MAN_01', 'Server');

  A complete script should connect to the AdminServer first:

  Connect('weblogic','password123','t3://localhost:7001');

  Start ('Cluster_Name', 'Cluster');

  Exit();

  If you need start the AdminServer as well, you can connect directly to the NodeManager and start it in nmStart.

  To stop the server/cluster, you can use the command "shutdown":

  stop ('SOACluster', 'Clusters', force = 'true');   -Add strength = "true" setting so that you do not wait for the present session of the user on this server.

  Or you can get the current status via command line:

  State ('SOACluster', 'Cluster');

  Once the sequence of commands you want in your WLST script, you can call it UNIX script as below:

  . /U01/Oracle/MW_HOME/wlserver_10.3/Server/bin/setWLSEnv.sh

  Java weblogic. WLST /u01/oracle/MW_HOME/wlserver_10.3/server/bin/startDomainScript.jy

• A query need to check whether the database should any recovery media...

  Hello
  
  I'm at the stage of editing a database and still be opened.at this point no matter what query is available or in any other way is there to check whether the database should any recovery media or Instance recovery... ?
  
  My goal is before opening the database, I make sure it's compatible or incompatible, and any form of recovery is necessary.
  
  Thank you and best regards,
  Rameau
  
  Published by: ragu.dba.in on 6 March 2013 12:57 AM

  Hi Lila,

  Use these queries, YVERT should match... no fuzzy file, don't check for recovering files in v$ recover_file... spend sufficient archives & realize the audits before open reset-logs

  set line 900
  set pages 900
  set echo on
  alter session set nls_date_format='dd-mon-yyyy hh24:mi:ss';
  set numwidth 13
  col CHECKPOINT_CHANGE# for 9999999999999999
  col CHANGE# for 9999999999999999
  column error format a10
  column member format a60
  show parameter db_recover
  select * from v$recover_file;
  select file#, status from v$datafile;
  select file#, checkpoint_change#, checkpoint_time, error from v$datafile_header;
  select distinct status from v$backup;
  select distinct  checkpoint_change#  from v$datafile_header;
  select hxfil file_id, fhscn scn, fhthr thread, fhrba_seq sequence, fhsta status from x$kcvfh;
  select distinct fhscn from x$kcvfh;
  select hxfil file_id, fhtnm tablespace_name from x$kcvfh;
  ##### Fuzzy Check ####
  alter session set nls_date_format = 'DD-MON-YYYY HH24:MI:SS';
  set feedback on
  set heading on
  set pagesize 1000
  set linesize 175
  column checkpoint_change# format 999999999999999999999
  select status, checkpoint_change#, fuzzy,
  to_char(checkpoint_time, 'DD-MON-YYYY HH24:MI:SS') as checkpoint_time,
  count(*)
  from v$datafile_header
  group by status, checkpoint_change#, fuzzy, checkpoint_time
  order by status, checkpoint_change#, fuzzy, checkpoint_time;
  

  Thank you
  Ajay more
  http://www.moreajays.com

• Must Oracle newspapers archived when retrieving the instance?

  Hi guys,.
  
  11g.
  
  Oracle must archived or journals just online during instance recovery? And why? The more detailed the better.
  
  1 million thanks!

  Oracle must archived or journals just online during instance recovery?

  And the docs say:

  Crash recovery requires that redo logs online: archived recovery logs is not necessary.

  http://docs.Oracle.com/CD/B10501_01/server.920/a96519/glossary.htm

  Good discussion at the link below:
  Journal archiving and recovery of Instance

  Concerning
  Girish Sharma