Clustering in Linux

As I know Oracle do provide a Failsafe Technology on Microsoft NT
Clustering Environment.
I want to know How can I do the same thing in Linux. OR How Can I
ensure that if my Oracle Database Server is down suddenly and I
can restart my database service with a minimum downtime. Is there
any solutions ?
null

You do not need to reinstall box 1.
Regarding the changes, they could be small if (and only if) the initial box was already prepared for RAC.
With this I mean
Do you have RAC software installed?
Are your datafiles already in a storage system that can be shared by both systems?
Have you already installed the datafiles on a supported cluster filesystem?
Do yo have an extra network card on the box for the comunication between the two nodes?
Installing a "supported" RAC is not put a box and plug two cables.
It is also the posibility of having also SAP enqueue replication configured ...
You can find some documentation on SAP on Oracle

Similar Messages

Weblogic 6.1 clustering on Linux

 Hi,
 I have clustered two managed servers (one on solaris machine and the other on
 linux red hat 7.1). When I try to boot the cluster, WLS start fine on solaris
 but on the linux machine, I am getting the following message:
 ####<Oct 5, 2001 11:51:12 AM EDT> <Emergency> <WebLogicServer> <simpson> <tuxmac
 hine> <ListenThread> <system> <> <000209> <ListenThread.run() failed: java.lang.
 UnsatisfiedLinkError: no weblogicunix1 in java.library.path>
 java.lang.UnsatisfiedLinkError: no weblogicunix1 in java.library.path
 at java.lang.ClassLoader.loadLibrary(ClassLoader.java:1312)
 at java.lang.Runtime.loadLibrary0(Runtime.java:749)
 at java.lang.System.loadLibrary(System.java:820)
 at weblogic.platform.Unix.<init>(Unix.java:14)
 at java.lang.Class.newInstance0(Native Method)
 at java.lang.Class.newInstance(Class.java:237)
 at weblogic.platform.OperatingSystem.getOS(OperatingSystem.java:75)
 at weblogic.t3.srvr.T3Srvr$2.run(T3Srvr.java:1004)
 at weblogic.t3.srvr.Callback.done(Callback.java:37)
 at weblogic.t3.srvr.ListenThread.newServerSocket(ListenThread.java:186)
 at weblogic.t3.srvr.ListenThread.run(ListenThread.java:223)
 ####<Oct 5, 2001 11:51:12 AM EDT> <Emergency> <WebLogicServer> <simpson> <tuxmac
 stening for connections. You should probably restart it.>
 Can somebody tell where the file weblogicunix1 is located? has this file been
 changed with WLS 6.1 on linux? If yes can you tell me the name of the new file
 so I can create the necessary link?
 Thanks again.
 Amen


 Using the generic file to install the software on linux does not install this automatically.
 The bea support send me this file in an email. I believe for the time being, the
 software is not package properly.
 Thanks for your help.
 Amen
 Kumar Allamraju <[email protected]> wrote:
 ><!doctype html public "-//w3c//dtd html 4.0 transitional//en">
 ><html>
 >You should find it uner $WL_HOME/wlserver6.1/lib/linux
 > directory
 >Amen Akakpo wrote:
 ><blockquote TYPE=CITE>Hi,
 > I have clustered two managed servers (one on solaris machine and
 >the
 >other on
 > linux red hat 7.1). When I try to boot the cluster, WLS start fine
 >on solaris
 > but on the linux machine, I am getting the following message:
 >####<Oct 5, 2001 11:51:12 AM EDT> <Emergency> <WebLogicServer>
 ><simpson> <tuxmac
 > hine> <ListenThread> <system> <> <000209> <ListenThread.run()
 >failed: java.lang.
 > UnsatisfiedLinkError: no weblogicunix1 in java.library.path>
 > java.lang.UnsatisfiedLinkError: no weblogicunix1 in java.library.path
 > at java.lang.ClassLoader.loadLibrary(ClassLoader.java:1312)
 > at java.lang.Runtime.loadLibrary0(Runtime.java:749)
 > at java.lang.System.loadLibrary(System.java:820)
 > at weblogic.platform.Unix.<init>(Unix.java:14)
 > at java.lang.Class.newInstance0(Native
 >Method)
 > at java.lang.Class.newInstance(Class.java:237)
 > at weblogic.platform.OperatingSystem.getOS(OperatingSystem.java:75)
 > at weblogic.t3.srvr.T3Srvr$2.run(T3Srvr.java:1004)
 > at weblogic.t3.srvr.Callback.done(Callback.java:37)
 > at weblogic.t3.srvr.ListenThread.newServerSocket(ListenThread.java:186)
 > at weblogic.t3.srvr.ListenThread.run(ListenThread.java:223)
 > ####<Oct 5, 2001 11:51:12 AM EDT> <Emergency> <WebLogicServer>
 ><simpson> <tuxmac
 > stening for connections. You should probably restart it.>
 >Can somebody tell where the file weblogicunix1 is located? has
 >this file been
 > changed with WLS 6.1 on linux? If yes can you tell me the name of
 >the
 >new file
 > so I can create the necessary link?
 >Thanks again.
 >Amen</blockquote>
 ></html>
 >

Oracle on Linux and Clustering

Are there any solutions for Oracle Clustering on Linux?
null

Hi,
the latest dev. kernel (2.3.12) contains raw device support. For
linux 2.2.10 there is a patch avail. from Stephen Tweedy. That
works great even with oracle8.
ftp://ftp.uk.linux.org/pub/linux/sct/fs/raw-io/
I guess it will be avail. in stable kernel soon.
Marcus
Douglas (guest) wrote:
: Linux does not support raw devices which are needed for OPS.
: Anyone fncy developing raw device support ?
: John Smiley (guest) wrote:
: : The Oracle product you need to support clusters for high
: : availability and scalability is Oracle Parallel Server
(OPS).
: : Unfortunately, the Linux Oracle8i documentation states that
OPS
: : is not yet available for Linux.
: : You could use a hot standby database as Freddy suggested, but
: : this does not require a clustered environment. Oracle 8i
: : (8.1.5) allows a hot standby database to be temporarily
placed
: : in an open, but read-only state. This is particularly useful
: : for batch reporting applications where the main database
needs
: : to be available 24x7 and the reporting load would degrade
: : performance of the main database. So, with Oracle8i, the hot
: : standby can be used as both a disaster recovery database, and
a
: : useful reporting database. With previous versions of Oracle,
: : the hot standby was not useable as anything but a disaster
: : recovery platform.
: : Freddy Wissing (guest) wrote:
: : : Volker Rath, DETECON (guest) wrote:
: : : : Are there any solutions for Oracle Clustering on Linux?
: : : I'm gonna begin experimenting with this.
: : : Oracle can do what's called a standby database, running on
a
: : : remote machine, possibly one in a linux cluster.
: : : The Linux HA project has come a long way, and I believe
could
: : : handle switching the standby database to active, and
assuming
: : the
: : : mac address, IP address and hostname, etc of the failed
node.
: : : Once I have time, I'm gonna try and put all this together.
: If
: : I
: : : have success, I'll be sure and post it.
: : : There are also commercial clustering solutions coming out
for
: : : Linux too. Either way, it won't be long.
: : : - Freddy Wissing
null

Making new oracle 10g cluster in redhat enterprise linux

Hy All
i have installed redhat enterprise server in 2 systems and both are attached with SANE network storage. i want to configure oracle 10 g clustering with this setup.
should i have to configure clustering in linux only or there is a feature of clustering in oracle?
if yes than how to do that? i dont have any idea about oracle
please somebody help me
with regards,
ali

Hi Ali!
Hardware clustering is differnet from Oracle clustering.
If you have the database you should investigate RAC http://www.oracle.com/pls/db102/to_toc?pathname=rac.102%2Fb14197%2Ftoc.htm&remark=portal+%28Administration%29
For the Application Server have a look at http://download-uk.oracle.com/docs/cd/B14099_18/core.1012/b14003/toc.htm
Hardware clustering with RedHat could be achieved with Piranha.
cu
Andreas

Oracle 8 Enterprise/Parallel Server availability on Linux

I have not downloaded the Linux port of Oracle 8 yet so please
bear with me... I would like to know whether Oracle Parallel
Server
is included in the distribution for Linux. Specifically, I
would like to build a system that has a failover node for a
client with
high-availability/reliability needs. Has anyone done this on
Linux yet?
I have looked at the Marketing blurb on Oracle Parallel Server
and it looks like OPS should suit my purpose. Any comments?
Caveats?
Thanks in advance,
Ty Haeber
Quetzal Systems Inc.
Vancouver, B.C.
null

Bud Alverson (guest) wrote:
: The Beowulf project supports clustering with Linux. Redhat
even
: has a release called "Extreme" that supports it in a fashion.
: In fact, Los Alamos has achieved the upper tier of the top 500
: fastest supercomputers in the world using Linux (#114, see:
: http://www.top500.org/top500.list.html). It's named Avalon and
: has 140 Alphas clustered via Redhat Linux 5.0. Read about it
at:
: http://loki-www.lanl.gov/papers/sc98/
To run Oracle Parallel Server (OPS) on a Linux-Cluster you would
need (Virtual) Shared Disks with Concurrent Access from all
Cluster-Nodes. As far as I know Beowulf does not provide
(Virtual) Shared Disks (a la VSD/RVSD on RS/6000 SP). NFS, etc.
will not work as substitute for Virtual Shared Disks since Raw
Partitions (Logical Volumes) with Concurrent Access are needed.
Maybe DB2 UDB EEE (a Shared-Nothing architecture) might become
available more easily as MPP DBMS on Linux, since it does not
need (Virtual) Shared Disks.
: Bud Alverson
: CrossNet Communications
: Michael Daskaloff (guest) wrote:
: : Ty haeber (guest) wrote:
: : : I have not downloaded the Linux port of Oracle 8 yet so
: please
: : : bear with me... I would like to know whether Oracle
Parallel
: : : Server
: : : is included in the distribution for Linux. Specifically, I
: : : would like to build a system that has a failover node for a
: : : client with
: : : high-availability/reliability needs. Has anyone done this
on
: : : Linux yet?
: : : I have looked at the Marketing blurb on Oracle Parallel
: Server
: : : and it looks like OPS should suit my purpose. Any
: comments?
: : : Caveats?
: : : Thanks in advance,
: : : Ty Haeber
: : : Quetzal Systems Inc.
: : : Vancouver, B.C.
: : Dear friend,
: : what I'd tell you might be wrong but you might consider my
: words.
: : I'm not very knoledgable about linux but I think there is
: : currently no hardware and software to support clusters on
: linux.
: : I'm sure Oracle Corp. will do the most of them to make I
: : Parallel Server (I think, yet there is no Enterprise Edition
: for
: : Linux) option for Linux as soon as there is such a
combination
: of
: : hardware/software.
: : Meanwhile if you really need I working and stable solution
: : go get Dec Alpha or RS/6000. There is already Parallel Server
: : for NT, which IMO currently runs only on Compaq hardware.
: : However I think You'd better you UNIX for your Parallel
Server
: : enviroment.
: : If you just need to reduce downtime You can you a feature of
: : Oracle called 'Standby Database' which is available for all
: : versions and Editions of Oracle8 (including that for Linux).
null

Converting Single Instance database into Oracle Fail-Safe on Linux

Hi All,
We have single instance Oracle10g database running on Linux RHEL4. We are looking to convert this database into Oracle Fail-Safe (Active-Passive). Does any one have document for Oracle fail-safe setup?
Regards,
Tushar

Tushar,
you can't use software for windows on Linux.
However, there's number of ways to implement high-availability solution on Linux Oracle.
You can go with Active-Active (RAC) - which means that you'll have two servers attached to the same database at the same time. You can use either, and if one goes down(due to power failure for instance), the second will still be available.
You can go with Active-Passive (Data Guard) - which means that you'll have one of the nodes being active and the second passive, but synchronized with the first one, so in case of failure of the first node you can activate the second one.
You can also go with linux clustering - which is most similar to MCS + fail-safe - you'll have two nodes clustered with linux clustering software (RH cluseter suite for instance) and connected to shared storage. One node is active, and in case of crash database instance will be automatically started on the second node. This is beneficial because you can use your resources more efficiently. In case if you have two databases you can run each of them on dedicated server and then, in case of node failure move to another one. However, the setup of such configuration is quite cumbersome.

Installing multiple instances accessing the same database

Hi,
I want to install two different instances of Oracle 10g in two different machines which will access the same database which will be stored in the shared storage.
Is it possible to install them without installing RAC? The instances will be one active and the other passive, so the services will be up in one server and down in the other and the switching (shutting down one server and starting up the other) will be manual.
Two servers will be running Linux and clustered in Linux level.
Does Oracle offers this solution without installing Clusterware software?
Thank you

> The instances will be one active and the other passive, so the services will be up in one server and down in
the other and the switching (shutting down one server and starting up the other) will be manual
Missed this part as I was thinking proper cluster and RAC.
This is neither. Yes, this can be done using two servers and shared storage.
Is it a good idea? Not really. As this configuration does not provide redundancy at physical database level. You loose that storage.. bye-bye database. Does not matter whether you have a 100 backup servers that can be used.
Thus the business reasons that you are trying to meet with this config have to be clarified and expectations determined.
Separate servers using Data Guard will be a far more superior solution in many respects.

Cisco CSS ICS via DWDM

We are currently splitting up a campus installation (2 datacenters with < 300m cable distance).
One datacenter remains on the campus, the other one is moved to another part of the town, approx. 30km away.
The two datacenters are interconnected using DWDM (don't have the exact specs at the moment, but I think we have got the equivalent of 16 duplexed 4Gb/s conenctions between the two data centers)
So far we have been able to move most of the equipment (including several members of Oracle RAC clusters on Linux and OpenVMS, VPN server farms, ESX cluster members and similar services), but we do not seem to bei able to get the Cisco CSS ICS link up on the DWDM.
Is there anything we can ask the DWDM provider to check, or is there no chance to get the ICS link up over DWDM?

Hi Martin,
I guess you are referring to ISC port, right?
As per CSS documentation: You must connect the ISC ports directly to the two CSSs. You cannot use Layer 2 devices on the ISC links between the two CSSs. Also, the ISC links must be dedicated to passing only ISC traffic.
For that reason I believe you need to reconsider your plan.
http://www.cisco.com/en/US/docs/app_ntwk_services/data_center_app_services/css11500series/v8.20_v8.10/configuration/redundancy/guide/ASR.html#wp1038263
Best regards,
Ahmad

Large table

Hi, I will have a large table with 1 million records inserted everyday. The table schema has a couple of CLOBs. We are going to keep 1 year's data, so the table will eventually have about 365 million records. I have following questions.
1) Is this amount of data can be handled by Oracle?
2) Is there anything I can do for performance tuning, like partition, etc.?
3) Any other advice for large table? Any recommended online documents/articles?
Thanks,
Zhe

Philippe Florent wrote:
You work with RAC and Linux, do you think this OS is mature to build an HA solution ? As many advices as people I asked. I don't expect a definitive answer of course, just thoughts...Well, I am a bit biased as I used Linux since the pre 1.0 kernel days. :-)
We have 5 RAC clusters - all running x86_64 h/w and Linux as the o/s. Linux today is equivalent today to any other nix system - from HP-UX to Solaris. Thus no backseat ito performance, flexibility and scalability. There seems to me to be a lot of FUD around Linux. Amongst the old nix hands, it is seen as a new snotty nose and immature kid on the block. This is helped along with marketing propaganda by companies like Microsoft.
Last year local Oracle asked me to do a Q&A session with one of their customers as they were looking at RAC and wanted answers from a user/customer perspective and not Oracle sales. Halfway through it we got to talking operating systems - and I was surprised to still find this idea that Linux is a hobbyist/immature type operating system and not comparable to a "real' o/s like Solaris or HP-UX for example.
There's a single fact that I use to try and counter this idea about Linux not being ready for running big commercial systems. The question: "What o/s do the vast majority of the 500 largest and fastest computer systems on this planet use?" The answer: <a href="http://www.top500.org/stats/list/35/osfam">Linux.</a>
455 of the world's faster 500 computer clusters use Linux. That is 91%. Up from around 84% last year this time. Not one of these run HP-UX. Only 5 runs Microsoft Windows. 2 runs Open Solaris. 19 runs AIX.
So if Linux is the predominant choice for o/s to run these large clusters, why then would it not be "+good enough+" for a (many times smaller) corporate cluster? In these large clusters, everything is magnified. Performance. Administration. Support. Flexibility. Robustness and stability. Costs. Scalability. Etc. And 91% of the time, Linux was selected to address these requirements and concerns.
So I like to reverse the question and instead ask "+What are your reasons for not using Linux and are these valid ones?+" :-)

Really strange cluster isolation problem....

We are still using Coherence 3.6.0 and is experiencing a really strange problem with two test clusters:
Cluster 1 uses wka a.b.c.d:9000 and cluster name "cluster1"
Cluster 2 uses wka e.f.g.h:9000 and cluster name "cluster2"
I can using JMX and log files see that both clusters have started successfully with node 1 (the single wka) using port 9000.
A client program first connects successfully to cluster 1 (using cluster 1 wka and cluster name as above). After the program terminate the same client program is run again this time versus cluster 2 (using cluster 2 wka and cluster name as above) but this time the connection is refused and to our total surprise the error message indicates that the Cluster 1 wka node a.b.c.d:9000 has refused connection and that the cluster name should be "Cluster 1"!!!!!!! We have checked and re-checked that the right combination of wka and cluster name (for cluster 2) is specified on the command line the second time....
Short of a DNS configuration problem (we are in fact using server names rather than physical IPs in the command line overrides) that would resolve the two wka node names to the same physical IP I cant see any solution (not involving woodo, alien intervention and/or the Bermuda triangle) to this behaviour. I actually tried pinging the two host names and could see that they indeed resolved to different IPs to eliminate this, unlikely but possible, problem...
It is my understanding that when specifying a unique wka for two clusters there is no need to set up unique multicast addresses / ports (since multicast is not at all used in this case) - can somebody confirm that this is indeed true or if this could have anything to do with our problem?!
Any suggestions or ideas (in addition to hiring a shaman to exorcise our servers :-)) are warmly appreciated...
/Magnus

After studying the problem more in detail I have come to the conclusion that this indeed seem to be a bug in Coherence 3.6.0. It does however not occur with 3.6.1 or 3.7.0 but if the problem is unknown I would have checked it out...
The problem is really easy to reproduce. Create two clusters (can be single storage node, local or remote does not matter) each with a unique WKA port (of course).
Execute a simple client program (non storage enabled) first against the one cluster then against the other. The program I used just looked up a cache ("near-test") and printed its size (in this case zero since I never loaded anything into the cluster).
You don't need to enable pof or anything specia. The cache-config iin the coherence.jar works fine. I have tested this both on distributed clusters on Linux and locally on a single Windows box.
You will notice that the program works against the first cluster but fails when run against the second.
/Magnus

Linux Clustering with oracle

Hai All,
Please forward the good pdf documents and links describing the Redhat Linux Clustering with Oracle.
Please help...
Shiju

Hai All,
Please forward the good pdf documents and
links describing the Redhat Linux Clustering with
Oracle.Check following link
http://www.oracle.com/pls/db102/to_pdf?pathname=install.102%2Fb15660.pdf&remark=portal+%28Getting+Started%29
Virag

DB+ASM with linux clustering

Dear All,
We are going to install Oracle single instance 11.2.0.3 with ASM, on linux clustering. Can anyone through some light on that. I searched in google and could not able to find good doc.
Can oracle binaries can be shared (with shared disk) between two servers and creating two listener will help ? (if node1 fails, db can be accessed by node2 listener).
Regards,
Anand.

AnandKumar wrote:
Is it possible to share the oracle binaries (i.e installing oracle in one server only) so that after failover we access the both the binaries and ASM data..Yes. Install the ORACLE_HOME on a clustered file system that can be mounted and used by another server. ASM supports ACFS. There are also OCFS and GFS. Never done this myself, but the Oracle installation docs cover this as far as I know.
Keep in mind though that the o/s hooks (such as an +/etc/inittab+ setting for the OHASD daemon) will not be included on a server that mounts and uses the cluster file system, on which an actual Oracle install/setup was not done. So it is not a mere matter of installing onto a clustered file system on one server and expect that to work when mounted by another server.

Linux clustering for 8i standard edition

Does Linux do clustering or fail-over for Oracle 8i STANDARD edition ?
I know for 8i ENTERPRISE edition, Oracle handles the clustering with the Parallel option. As well, for 8i Standard Edition, Oracle offers 'fail-safe' for the NT platform but not for LINUX.
If anyone knows, please advise. TIA.

There is no Paraller Server for Linux now.
But there is possibilities for HA
cluster :)
For instance, http://oss.sgi.com/failsafe
(or be somethig like this), www.missioncriticallinux.com
Then, there is several non open source
software. www.steeleye.com,
if I rememember.
As I know Oracle plan Parallel server for Linux.

Linux clusters and oracle

Hello gentlemen!
Does anybody have experience with linux clusters and oracle?
Does this combination exist? Is linux great in this role?
Any links on this topic?
Any help would be greatly appreciated.
Best regards,
San.
null

This contradicts the information in the Oracle® Database Release Notes 10g Release 2 (10.2) for Linux x86-64:
In Oracle Database Oracle Clusterware and Oracle Real Application Clusters Installation Guide, Chapter 2, "Preinstallation," in the section "Oracle Clusterware Home Directory," it incorrectly lists the path /u01/app/oracle/product/crs as a possible Oracle Clusterware home (or CRS home) path. This is incorrect. A default Oracle base path is /u01/app/oracle, and the Oracle Clusterware home must never be a subdirectory of the Oracle base directory.
A possible CRS home directory is in a path outside of the Oracle base directory. for example, if the Oracle base directory is u01/app/oracle, then the CRS home can be an option similar to one of the following:
u01/crs/
/u01/crs/oracle/product/10/crs
/crs/home
This issue is tracked with Oracle bug 5843155.

Gentoo Linux and Microsoft Failover Clusters / Hyper-V

Hello,
Hoping there are a few people on the boards familiar with running Gentoo Linux guests under Microsoft FailOver Cluster / Hyper-V hosts.
I have four Gentoo Linux guest VMs (running kernel 3.12.21-r1) running under the Microsoft Failover Cluster system with Hyper-V as the host. All of the Hyper-V drivers are built into the kernel (including the utilities and balloon drivers) and generally they
run without issue.
For several months now, however, I have been having strange issues with them. Essentially they stop responding to network requests after random intervals. However, these intervals aren't a few minutes or hours from each other; more like days or even weeks before
one of them will stop responding on the network side.
The funny thing is that the VMs themselves on the console side still responds. However, if I issue a reboot command on the externally non-responsive VM, the system will eventually get to a stage where all of the services are stopped and then hangs right after
the "mounting remaining system ro" line (or something like that).
The Failover Cluster Manager then reports that the system is "Stopping" but the system never reboots.
I have to completely restart the HOST system so that either (A) the VM in question transfers to another host and starts responding again or (B) when the HOST comes back up I can work with the VM again.
This *ONLY* happens on the Gentoo Linux guest VMs and not my Windows VMs.
Wondering if anyone has hints on this.
Thank you for your time.
Regards, Christopher K.

Hi ckoeber,
Hyper-V has support most of Linux distribution but not all because it’s have so many distribution, until now the Gentoo Linux not in the support list, you can refer the following
KB to know more about the detail:
Linux Virtual Machines on Hyper-V
http://technet.microsoft.com/en-US/library/dn531030
Hope this helps.
We
are trying to better understand customer views on social support experience, so your participation in this
interview project would be greatly appreciated if you have time.
Thanks for helping make community forums a great place.

Clustering in Linux

Similar Messages

Maybe you are looking for