Let’s say you have your Oracle DB backups on Ashburn and for DR requirements you must have a copy of those files on object storage at Phoenix.
Is it possible to do it ?
Yes, it is but OCI doesn’t have a native replication solution like AWS S3 and it also does not allow you to run bulk copy.
Exactly, you have to copy object by object to the DR region at this time.
Which is very simple task when using oci cli and bash. A code example below:
#!/bin/bash
export OCIHOME=/home/oracle/ocicli
export OBUCKET=ashbucket
export DBUCKET=phobucket
export NAMESPACE=namespace
# object store list
$OCIHOME/oci os object list --bucket-name=$OBUCKET --all | grep name | awk '{ print $2 }' | sed 's/"//g' | sed 's/,//g' > /tmp/ostore.list
#sintaxe:
#oci os object copy --namespace-name <object_storage_namespace> --bucket-name <source_bucket_name> --source-object-name <source_object> --destination-namespace #<destination_namespace_string> --destination-region <destination_region> --destination-bucket <destination_bucket_name> --destination-object-name <destination_object_name>
# copy to DR region
for id in `cat /tmp/ostore.list`
do $OCIHOME/oci os object copy --namespace-name=$NAMESPACE --bucket-name=$OBUCKET --source-object-name=$id --destination-namespace=$NAMESPACE --destination-region=us-phoenix-1 --destination-bucket=$DBUCKET
done
Object Storage is a regional service so you must authorize the Object Storage service for each region carrying out copy operations on your behalf.
In this example, you have to authorize the Object Storage service in region US East (Ashburn) to manage objects on your behalf.
To accomplish this, create the following a policy:
Allow service objectstorage-us-ashburn-1 to manage object-family in tenancy
Hopefully OCI will soon deploy a native replication capability 🙂
For more information, Copying Objects documentation is here.
Create a group to your and assign proper privileges like:
Allow group StorageAdmins to manage file-family in compartment PROD
Schedule this shell script to run on a regular internal that fits your needs.
But in my case I had to keep a limited amount of backups based on same information such as: environment (prod or dev) and customer retention policy (bronze, silver or gold).
So I wrote another simple shell to accomplish this: fscleanup.sh
!/bin/bash
export CI=/home/oracle/ocicli
export FSSOCID=<file storage ocid>
export TSTAMP=`date +%Y%m%d`
export KEEP=42
# dump backups to tempfile
$CI/oci fs snapshot list --file-system-id=$FSSOCID | grep '"id"' | awk '{print $2}' | sed 's/"//g' | sed 's/,//g' > /tmp/fss.bkp
#count
CT=`cat /tmp/fss.bkp | wc -l`
#remove backups older then $KEEP
if [ "$CT" -gt $KEEP ]; then
DIFF=$(expr $CT - $KEEP)
for id in `tail -$DIFF /tmp/fss.bkp`
do
$CI/oci fs snapshot delete --force --snapshot-id $id
done
else
echo "less then KEEP: $KEEP"
fi
Kdump is the Linux kernel crash-dump mechanism and Oracle recommends that you enabled it. You can configure Kdump using system-config-kdump utility.
So, if you are running Oracle Enterprise Linux UEK kernel and have faced a kernel panic, follow the steps below to install crash utility and analyse the vmcore file.
To use crash, packages debuginfo and debuginfo-common must also be installed.
You will find vmcore files under /var/crash/<127.0.0.1-TIMESTAMP>
Output example:
[root@machine 127.0.0.1-2020-01-26-20:29:08]# crash vmcore /usr/lib/debug/lib/modules/4.1.12-124.35.2.el7uek.x86_64/vmlinux
crash 7.2.3-10.el7
Copyright (C) 2002-2017 Red Hat, Inc.
Copyright (C) 2004, 2005, 2006, 2010 IBM Corporation
Copyright (C) 1999-2006 Hewlett-Packard Co
Copyright (C) 2005, 2006, 2011, 2012 Fujitsu Limited
Copyright (C) 2006, 2007 VA Linux Systems Japan K.K.
Copyright (C) 2005, 2011 NEC Corporation
Copyright (C) 1999, 2002, 2007 Silicon Graphics, Inc.
Copyright (C) 1999, 2000, 2001, 2002 Mission Critical Linux, Inc.
This program is free software, covered by the GNU General Public License,
and you are welcome to change it and/or distribute copies of it under
certain conditions. Enter "help copying" to see the conditions.
This program has absolutely no warranty. Enter "help warranty" for details.
GNU gdb (GDB) 7.6
Copyright (C) 2013 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later http://gnu.org/licenses/gpl.html
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law. Type "show copying"
and "show warranty" for details.
This GDB was configured as "x86_64-unknown-linux-gnu"…
KERNEL: /usr/lib/debug/lib/modules/4.1.12-124.35.2.el7uek.x86_64/vmlinux
DUMPFILE: vmcore [PARTIAL DUMP]
LOAD AVERAGE: 0.88, 0.26, 0.18
TASKS: 366
NODENAME: machine
RELEASE: 4.1.12-124.35.2.el7uek.x86_64
VERSION: #2 SMP Mon Jan 13 16:42:30 PST 2020
MACHINE: x86_64 (3400 Mhz)
MEMORY: 15.7 GB
PANIC: "BUG: unable to handle kernel NULL pointer dereference at 0000000000000030"
PID: 24893
COMMAND: "oracle_24893_ic"
TASK: ffff880168877000 [THREAD_INFO: ffff8801a6eb8000]
CPU: 0
STATE: TASK_RUNNING (PANIC)
Great. Now the interesting part: let’s try to launch a DB system on each VCN.
You have to inform a lot of parameters so I’m using a json file.
Attempt to launch a DB on VCN1 fails with error below:
oci db system launch --from-json file://db_19c_vcn1.json ServiceError: { "code": "InvalidParameter", "message": "domain name cannot be an empty string.", "opc-request-id": "6B793936B1BB4D33BC0DDE6399C21B8B/6049601F5F54B75274E10E48542D39AD/D048ADC38DCDE42EA972F2D3E65FB9AE", "status": 400 }
Now on VCN2, db launch works ! (output truncated):
Bottom line is: Don’t forget to define DNS while using oci cli for VCN provisioning. It is not required by default but you may miss this later and it is not possible to change DNS after VCN creation.
CONFIGURE DEVICE TYPE disk PARALLELISM 2;
CONFIGURE DEVICE TYPE sbt PARALLELISM 2;
RMAN always allocates the number of channels specified in PARALLELISM, using specifically configured channels if you have configured them and generic channels if you have not. Note that if you configure specific channels with numbers higher than the parallelism setting, RMAN will not use these channels.
$ORACLE_HOME/bin/dbupgrade -n 4
-n options specifies the number of processes to use for parallel operations.
Non-CDBs: The -n parameter specifies the number of SQL processes to use when upgrading the database.
Multitenant architecture databases (CDBs): The number of PDBs upgraded concurrently is controlled by the value of the -n parameter. Multiple PDB upgrades are processed together. Starting in Oracle Database 12c, the default value for multitenant architecture databases is the number of CPUs on your system. A cpu_count equal to 24 equates to a default value of 24 for -n.
Values for the -n parameter:
Non-CDBs: The maximum value for -n is 8. The minimum value is 1. The default value is 4.
Multitenant architecture databases (CDBs): The maximum value for -n is unlimited. The minimum value is 4. The maximum PDB upgrades running concurrently is the value of -n divided by the value of -N.
-N option specifies the number of SQL processors to use when upgrading PDBs.
For non-CDBs, this parameter is ignored.
For CDBs, the maximum value is 8. The minimum value is 1. The default value is 2.
So without specifying values for the parameters -n and -N (that is, accept the default value of -N, which is 2, and accept the default value of -n as the CPU_COUNT parameter value, which is 24).
The following parallel processing occurs:
12 PDBs are upgraded in parallel (CPU_COUNT divided by 2)
2 parallel processes run for each PDB
That is what came from the top of my head so it should have more parallel options 🙂
Thanks Tim to get the community together every year ! #ThanksOGB
Na minha opinião o evento foi, em todos os aspectos, melhor do que no ano passado. A organização leva muito a sério a questão de melhoria contínua e por isso é muito importante que todos preencham a ficha de avaliação para que no próximo ano o evento seja melhor ainda !
A localização, as instalações, os palestrantes e a audiência foram incríveis !
Em todas as palestras que assisti (e também na que ministrei) a sala estava cheia e com gente interessada em aprender e trocar experiências.
Tivemos Product Managers da Oracle trazendo conteúdo exclusivo e super disponíveis para tirar dúvidas. Eu mesmo fiz um monte de perguntas para o Ricardo Gonzalez (PM para Oracle Cloud Migration, ACFS & RHP) e acabei ganhando uma caneca do evento 🙂
A última palestra foi do Alex Gorbachev (CTO da Pythian) sobre a evolução da carreira dele e como ele enxerga o futuro do DBA Oracle. Sim, vamos ter que mudar/evoluir muito !
Em resumo: o melhor evento de tecnologia Oracle do país … melhor que o Oracle Open World Brasil de longe.
Perdeu o evento desse ano ? Faça o cadastro gratuito no site guob.com.br e veja as palestras disponíveis.
I hit error ORA-19563 during rman restore last week.
I was restoring a database backup from Oracle 11.2 on ASM to a Oracle 11.2 on Filesystem.
Although not the same procedure, MOS RMAN Duplicate Fails With RMAN-06136 ORA-19563 (Doc ID 453123.1) helped me on this issue when it mentioned I could have duplicate filenames.
There was duplicated filenames indeed !
My restore script was:
run {
set newname for database to '/oradata/%b';
restore database;
switch datafile all;
}
Which means that datafile 55 was overwritten by datafile 56 hence rman error on switch step.
Quick fix:
remove datafile 56
restore only datafile 55 and 56 using set newname clause but specifying different filename:
run {
set newname for datafile 55 to '/oradata/svm910t02.dbf';
set newname for datafile 56 to '/oradata/svm910t03.dbf';
restore datafile 55;
restore datafile 56;
}
run original script again which this time will just perform the switch step
How to prevent this from happening again ?
Check if you have duplicate filenames before restoring:
select substr ( file_name, instr( file_name, '/', -1)) file_name, count() from dba_data_files group by substr ( file_name,instr( file_name, '/', -1)) having count() > 1
/
If yes, run rman restore like this:
run {
set newname for datafile 56 to '/oradata/svm910t03.dbf';
set newname for database to '/oradata/%b';
restore database;
switch datafile all;
}
More information can be found at Oracle 11.2 docs here.
