How Load Balance works in RAC?

Hi All,
I have a 2-node 10gR2 RAC on RHEL4. One of our web based application is connecting to the RAC database. Today, while starting up the application, one of our node(node1) was down and all the connections from the app are connected to the DB using other node(node2). There were around 120 connection from application. Again, i brought back the node1 but still, the connections were on node2 and no connection on node1.
My doubt is, how the load balancing works in RAC? is there limit like after some number of connections, it will load balance to other node?
Thanks,
Praveen.

If all the connections go to node 2 cause node 1 is down, then, when node 1 comes back, those connections already established on node 2 will remain on node 2. There is no mechanism to migrate existing connections. However, over time, as more connections come in, they should go to node 1. Over time, the load should balance out. But, the only way to immediately rebalance would be to force existing connections to recycle their connection.
If one node is capable of handling the load, I wouldn't worry about it....over time, things will balance out.
-Mark

Similar Messages

Question on how does load balancing work on Firewall Services Module (FWSM)

Hi everyone,
I have a question about the algorithm of load balancing on Firewall Services Module (FWSM).
I understand that the FWSM supports up to three equal cost routes on the same interface for load balancing.
Please see a lower simple figure.
outside inside
--- L3 SW --+
|
MHSRP +--- FWSM ----
|
--- L3 SW --+
I am going to configure the following default routes on FWSM point to each MHSRP VIP (192.168.13.29 and 192.168.13.30) for load balancing.
route outside_1 0.0.0.0 0.0.0.0 192.168.13.29 1
route outside_1 0.0.0.0 0.0.0.0 192.168.13.30 1　　　　　　
However I don't know how load balancing work on FWSM.
On FWSM, load balancing work based on
Per-Destination ?
Per-Source ?
Per-Packet ?
or
Other criteria ?
Your information would be greatly appreciated.
Best Regards,

Configuring "tunnel default gateway' on the concentrator allowed traffic to flow as desired through the FWSM.
FWSM is not capable of performing policy based routing, the additional static routes for the VPN load balancing caused half of the packets to be lost. As a result, it appears that the VPN concentrators will not be able to load balance.

Load balancing in 10g RAC

why we need to have load balancing(client and server side load balancing) in 10g RAC environment

853910 wrote:
why we need to have load balancing(client and server side load balancing) in 10g RAC environmentIn simple words because you don't want just one node to be getting busy with all the workload coming up but you do want the work to be divided among all the nodes in teh cluster since the RAC is based on the principle of divide and rule .
Aman....

Functional load balancing in oracle RAC

Hi all,
Could anyone suggest how to setup functional load balancing in oracle RAC? Will there be any performance improvement in doing so? I implemented server side load balancing and the performance of INSERT, UPDATE and DELETE are very poor. Will functional load balancing be of any use? Please tell me the steps of how to implement functional load balancing?
Regards,
Aravind
Edited by: Aravind K R on Jul 12, 2011 2:49 PM

Aravind K R wrote:
Hi all,
Could anyone sujjest how to setup functional load balancing in oracle RAC? Will there be any performance improvement in doing so? I implemented server side load balancing and the performance of INSERT, UPDATE and DELETE are very poor. Will functional load balancing be of any use? Please tell me the steps of how to implement functional load balancing?
Regards,
Aravind
Hi,
Load balancing will not necessary increase performance, it just distribute the load more evenly.
Is the DML statement only slow in specific instance?
To prove if there is problem with load balancing, have you tried connect to one instance and perform same DML statement? compare the result
Cheers

How the balance works for http/https in CSS?

I want to understand how the balance works for http/https in the CSS.
As per doc,
- http/1.0=single URL per TCP connection
- http/1.1=persistent connection.
Q1. my understanding is http is stateless connection so needs TCP session for each URL. how the http/1.1 works with persistent. keep the same TCP session for a multiple URL request?
Q2. https is using single URL and continue processing until terminate the https instead of open another URL in the middle of transaction. in this case, I think the client can stick into same service based on the assumption the CSS support persistent. if then, no advanced sticky(ex, srcip) required?
Q3. looks below both are analogy. what's difference between them?
- balance srcip(same src IP to the same service)
- advanced-balance sticky-srcip
Q4. what's balance decision mechanism for "balance roundrobin" to distribute evenly? ex, in case of multiple URL request coming from same client. evenly distribute URLs?
Regards,

Hello,
first let me clarify 1 point.
HTTP/HTTPS are standards that are defined in RFC.
For HTTP/1.1 you can check the following RFC
http://www.faqs.org/rfcs/rfc2616.html
Therefore, the behavior of HTTP 1.1 is not defined by the CSS.
Q1- HTTP/1.1 simply keeps 1 TCP connection to send a received mutiplie HTTP request/response.
HTTP/1.0 will open 1 TCP connection for every HTTP request.
Q2- HTTPS is just HTTP over SSL.
So basically the same rule as above applies.
HTTP/1.1 can use 1 SSL connection for many HTTP request/response while HTTP/1.0 will use 1 SSL connection for each HTTP request.
Therefore, if you have customer using HTTP/1.0 you need some form of stickyness to guarantee that every connections will go to the same server.
Even if only using HTTP/1.1 you may need stickyness.
A user could disconnect and reconnect and require to be loadbalanced to the same server as before.
Q3- There is a big different between balance srcip and sticky-srcip.
The balance srcip simply hash the source ip address to find the destination server.
The problem of this method is that the loadbalancing is not guarantee to be evenly distributed between the servers.
With sticky-srcip, you use a normal balance method like round-robin, and then you create a sticky entry in a sticky table.
Next time this user comes back we first check the sticky entry to find the destination server.
The advantage is that it guarantess your users will be evenly distributed among the server.
Q4- roundrobin is applied to connection - not url [by default].
So if you have 2 users and they both open 1 connection, the CSS will send 1 connection to 2 different servers.
So each server has 1 connection.
If one user sens 10 URL and the other 1 sends only 1, one server will have 10 url to process while the other only one.
That's if you are using HTTP/1.1 and use persistent mode on the CSS.
You can break persistency and split the url.
I run out of space and time to explain you everything.
I suggest you go read the RFC or a book on HTTP.
Also read the CSS configuration guide.
There is much more you need to know if you want to take full advantage of the CSS like cookie, ssl offloading, L7 rules vs L3/4 rules, ...
Regards,
Gilles.

Who takes care of load balancing work in weblogic cluster ?

Hi Folks,
How load balancing is handled internally by weblogic cluster ? Does Admin server takes care of cluster load balancing ? According to me Admin has nothing to do with load balancing in cluster . Manage server continue to run even when Admin server is down.
I am not looking for explanation about different load balancing algorithm. I am interested in who executes those algorithms in weblogic cluster and how .
Cheers
B.

Hi,
You need to configure loadbalancer and there are different ways to do this. Please go through the below thread to user Apache Loadbalancer.
problem with Apache Proxy plugin
Regards
Suresh.

HOW WISM LOAD BALANCING WORKS ?

We currently have a campus with 4 WISM on 4 6509 .
After a reload of one of these WISM, all the AP ( AIR-AP1231 ) disappear from this WISM :
(WiSM-slot1-2) >show ap summary
Number of APs.................................... 1
Before the reload this WISM has 80 AP .
The other 79 AP are now registered with another WISM .
Is this normal or ( load balancing ) or what ?
Any help is apreciated

You have to check the primary/secondary/tertiary settings on the WLC, but for each access point. Go to the Wireless tab and click on an access point for this setting.
AP Fallback mode is on the WLC under the Controller tab. It forces an access point to join its primary controller if possible. If it's disabled, the access point will be happy to stay on a non-primary controller until it reboots. If it's enabled, then as soon as its primary becomes available (either because it comes up or because you specify a new primary), it will join that controller.
AP Fallback is convenient for load-balancing, but it can cause significant downtime if a controller goes down and comes back up immediately. All your APs will fail over to another controller at the initial drop, but they'll all switch back over once the original controller comes back. Each time will cause clients to drop. If a controller starts flapping for any reason, your wireless network will be seriously disrupted. Of course, if you leave it disabled and a controller goes down, all access points will hop over to their secondary controller and never go back. This isn't so bad, though, because you can just manually set it for AP fallback at a convenient time, let them all switch back, and then disable fallback again.
If you do not active AP fallback mode, be sure to reset each AP after you give it a primary/secondary/tertiary controller. Otherwise it will sit tight on the current controller.
Also, make sure that the Master Controller mode is disabled. This forces all access points to join it if you have no primary controller listed. It's under the Controller tab, under Advanced.

Load balancing in Oracle RAC 11g (11.1.0.6.0)

Hi!
One of our customers use load balancing, now all users get connected against node 2.
--The configuration of load balancing will be correct I think, have worked before
--How do I verify the load on each node ?
--What can be wrong?
--Need help, what to check?
--tnsnames.ora looks like
X.world =
(DESCRIPTION =
(ADDRESS = (PROTOCOL = TCP)(HOST = X)(PORT = 1521))
(ADDRESS = (PROTOCOL = TCP)(HOST = X)(PORT = 1521))
(LOAD_BALANCE = yes)
(CONNECT_DATA =
(SERVER = DEDICATED)
(SERVICE_NAME = X.X.se)
Regards
T

Hi,
--The configuration of load balancing will be correct I think, have worked before
--How do I verify the load on each node ?
--What can be wrong?
--Need help, what to check?
Services
Traditionally an Oracle database provided a single service and all users connected to the same service.
A database will always have this default database service that is the database name.This service cannot be modified and will always allow you to connect to the database.
Do not use the default database service, create at least one service.
With Oracle Database 11g, a database can have many services (up to a maximum of 115 per database).
The services hide the complexity of the cluster from the client by providing a single system image for managing work. Applications and mid-tier connection pools select a service by using the Service Name in their connection data.
http://www.oracle.com/technetwork/database/clustering/overview/awmrac11g-133673.pdf
http://download.oracle.com/docs/cd/B28359_01/rac.111/b28254/hafeats.htm
Regards,
Levi Pereira

Load balancing of DB RAC using weblogic server

Hi All,
I want to test the load balancing of Data base RAC for my application. I got the box which is installed RAC in it with two nodes. I configured the multidatasource which points to individual data sources to each node.
Is this configuration is correct or not? to test the load balancing of the Database. I come to know that there is something called SCAN from oracle 11g but it is not supported 100% by web-logic server(we are using weblogic server 9.2MP3).
How many requests I made all the requests are going to the Node1 only not even one is gone to the node2. Any specific reason.
Thanks,
Venkat.

Another alternative is, keep the property file in any pre-defined location. Write a class which reads the properties from the file and returns the one which is requested by caller and deploy this class. Whenever you have to change the properties just update the property file on server and next call to fetch the property should return the updated one.
Downside of this approach is file I/O everytime. To overcome that you can actually "cache" the properties in a hashmap. Basically when any property if requested, first check the hashmap, if not found then only read from property file and also update in hash map. Next time same property will be returned from hash map itself. The hash map will be cleared at every server restart since its in the memory. You will also need to build a method to clear the hashmap when you update the values in the property file on server.
This solution would be suitable for small size files and when network overhead of calling a DB needs to be avoided.

JDBC load balancing with non RAC replicated instances

Hi,
The thin JDBC load balancing feature only works with RAC clusters or it can be used with two non RAC replicated instances using the same service name?
Thanks,
Ezio.

Thank you Douglas for your time.
I assumed the first case (database connection load balancing). I thought that the listener would do the load balancing and failover trick. So, I configured the JDBC connection as follow:
String MM_loadBalance1_DRIVER = "oracle.jdbc.driver.OracleDriver";
String MM_loadBalance1_USERNAME = "scott";
String MM_loadBalance1_PASSWORD = "tiger";
String MM_loadBalance1_STRING = "jdbc:oracle:thin:@(DESCRIPTION=" +
" (LOAD_BALANCE=on) (FAILOVER=on)"+
" (ADDRESS_LIST="+
" (ADDRESS=(PROTOCOL=TCP)(HOST=test1) (PORT=1521))"+
" (ADDRESS=(PROTOCOL=TCP)(HOST=test2)(PORT=1521)))"+
" (CONNECT_DATA=(SERVICE_NAME=test)))";
Driver DriverRecordset1 = (Driver)Class.forName(MM_loadBalance1_DRIVER).newInstance();
Connection ConnRecordset1 = DriverManager.getConnection(MM_loadBalance1_STRING,MM_loadBalance1_USERNAME,MM_loadBalance1_PASSWORD);
But, when one of the instances was down (test1), the connection failed, it didn't failed over to the other instance (test2).
So I wondered if it failed due to some misconfiguration issue or because it only works with RAC instances...

Automatic load balance for oracle RAC

Hi there,
we have application, that use local oracle 9i database and every users use terminal application. That meaning, users do connect to hostname via terminal and invoke application.
We want migrate to RAC with cluster composed with 2 nodes. On every node must be
unique oracle_sid, but database is same on disk array.
My question is how configure users terminal profile which be comensatory automatic load balance .
Thanks

Hi,
This software simulates a real environment.
It creates a schema with tables, sequence, index, functions, etc..
Preferably in a client host it simulates several users (configurable) connecting to the database.
Then the test is simulated like in the real environment.
Check this:
RAC speedup example
This demo shows an example of RAC speedup
http://apex.oracle.com/pls/apex/f?p=44785:24:1688430074669391::NO:24:P24_CONTENT_ID,P24_PREV_PAGE:4946,2
RAC scale example
This demo shows an example of how a 2 node RAC database can scale in performance.
http://apex.oracle.com/pls/apex/f?p=44785:24:1688430074669391::NO:24:P24_CONTENT_ID,P24_PREV_PAGE:4945,2
Regards,
Levi Pereira

Load Balancing with 10g RAC

Hi,
I have one query regarding balancing load with RAC.
We have some batch processes executed from shell script in Oracle RAC environment where a separate node is used for failover only. There are some issues related with performance. We are looking for a solution to utilize 'idle' node used for failover to resolve this issue.
I want to know whether the below mentioned scenario is possible to achieve with 10gRel2 RAC.
A batch process is running and it is utilizing all resources (CPU) of instance 1, then it is possible by any means that this process be split and distributed to idle instance 2 on another node?
Looking for expert advice.
Regards
Amit

Hi Amit,
I;ve done lots of RAC implementations, and it;s critical to set-up the primary node accordig to the "type" of table access. For example, cistomer queries on node 1, order queries on node 2, etc. Minimizing pinging is very important!
Staring in Oracle 10g release 2, Oracle JDBC and ODP.NET provide connection pool load balancing facilities through integration with the new “load balancing advisory” tool. This replaces the more-cumbersome listener-based load balancing technique.
I have more notes here:
http://www.dba-oracle.com/oracle10g_tuning/t_grid_tuning.htm
Hope this helps. . .
Donald K. Burleson
Oracle Press author

Load Balancing issue on RAC setup

1. We have a Oracle RAC cluster having two node say node1-vip and node2-vip
2. We have two client software (both uses weblogic as app server) say ANP and HDM
3. We configure connect string for HDM by giving node1-vip as primary and node2-vip as secondary. Here is the connect string we provide
jdbc:oracle:thin:@(DESCRIPTION=(ADDRESS_LIST=(ADDRESS=(PROTOCOL=TCP)(HOST=node1-vip)(PORT=1521))(ADDRESS=(PROTOCOL=TCP)(HOST=node2-vip)(PORT=1521))(FAILOVER=on)(LOAD_BALANCE=off))(CONNECT_DATA=(SERVER=DEDICATED)(SERVICE_NAME= anpdb)))
4. We configure connect string for ANP by giving node2-vip as primary and node1-vip as secondary. Here is the connect string we provide
jdbc:oracle:thin:@(DESCRIPTION=(ADDRESS_LIST=(ADDRESS=(PROTOCOL=TCP)(HOST=node2-vip)(PORT=1521))(ADDRESS=(PROTOCOL=TCP)(HOST=node1-vip)(PORT=1521))(FAILOVER=on)(LOAD_BALANCE=off))(CONNECT_DATA=(SERVER=DEDICATED)(SERVICE_NAME= anpdb)))
We are observing that connection for HDM is only going to node1-vip however connection for ANP sometime goes to only node2-vip however sometime it is getting distributed between node1-vip and node2-vip. Whenever there is more connection on node1-vip than node2-vip then connection for ANP is only going to node2-vip however whenever number of connection at node2-vip is equal or greater than node-1 vip, connection for ANP gets distributed between node1-vip and node2-vip
Thanks
Sumit

Hi,
jdbc:oracle:thin:@(DESCRIPTION=(ADDRESS_LIST=(ADDRESS=(PROTOCOL=TCP)(HOST=node1-vip)(PORT=1521))(ADDRESS=(PROTOCOL=TCP)(HOST=node2-vip)(PORT=1521))(FAILOVER=on)(LOAD_BALANCE=off))(CONNECT_DATA=(SERVER=DEDICATED)(SERVICE_NAME= anpdb)))
Why have you put *(LOAD_BALANCE=off)* ?
And from your post, you are not trying to use actual load balancing.
You are trying to give connectivity like this HDM->node 1 and ANM->node 2
In that case you need tp specify only one server vip address.
When you configure the connection entries like above, when a new user session connection is requested, listener will connect the user to whichever address is reachable at that time.
Hope you got my point...
Regards,
Mahesh.

Load Balancing not load balancing!

Hi,
We have inherited a 6.3 2005 Q1 installation of Access Manager that has the following problem.
All logons appear to be load balanced correctly however all Policy Agent requests are expressing a preference for one or other node at a time.
For example, during the last 4 days all PA traffic has been hitting Srv01. This leads to an 80/20 balance. This morning on Srv01 an error in the amSession logs appeared on Srv01, "SessionRequestHandler NullPointerException" and from this point on Srv02 started to take the Policy Agent requests. So now the traffic is still 80/20 but wth Srv02 handling the 'lion's share'.
Can anyone point me to documentation that describes in detail how Load Balancing working in the products?
Thanks,
EddieT

Hi,
Sorry I should have been more specific. The error we are seeing has been reported before but no resolution posted.
Thanks.
29/05/2009 08:57:05:119 AM IST: Thread[service-j2ee,11,main]
ERROR: SessionRequestHandler encounterd exception
com.iplanet.sso.SSOException: Session was not obtained.
at com.iplanet.sso.providers.dpro.SSOProviderImpl.createSSOToken(SSOProviderImpl.java:177)
at com.iplanet.sso.SSOTokenManager.createSSOToken(SSOTokenManager.java:305)
at com.sun.identity.session.util.RestrictedTokenContext.unmarshal(RestrictedTokenContext.java:125)
at com.iplanet.dpro.session.service.SessionRequestHandler.processRequest(SessionRequestHandler.java:139)
at com.iplanet.dpro.session.service.SessionRequestHandler.process(SessionRequestHandler.java:112)
at com.iplanet.services.comm.server.PLLRequestServlet.handleRequest(PLLRequestServlet.java:195)
at com.iplanet.services.comm.server.PLLRequestServlet.doPost(PLLRequestServlet.java:147)
at javax.servlet.http.HttpServlet.service(HttpServlet.java:767)
at javax.servlet.http.HttpServlet.service(HttpServlet.java:860)
at sun.reflect.GeneratedMethodAccessor203.invoke(Unknown Source)
at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:25)
at java.lang.reflect.Method.invoke(Method.java:585)
at org.apache.catalina.security.SecurityUtil$1.run(SecurityUtil.java:249)
at java.security.AccessController.doPrivileged(Native Method)
at javax.security.auth.Subject.doAsPrivileged(Subject.java:517)
at org.apache.catalina.security.SecurityUtil.execute(SecurityUtil.java:282)
at org.apache.catalina.security.SecurityUtil.doAsPrivilege(SecurityUtil.java:165)
at org.apache.catalina.core.ApplicationFilterChain.internalDoFilter(ApplicationFilterChain.java:257)
at org.apache.catalina.core.ApplicationFilterChain.access$000(ApplicationFilterChain.java:55)
at org.apache.catalina.core.ApplicationFilterChain$1.run(ApplicationFilterChain.java:161)
at java.security.AccessController.doPrivileged(Native Method)
at org.apache.catalina.core.ApplicationFilterChain.doFilter(ApplicationFilterChain.java:157)
at org.apache.catalina.core.StandardWrapperValve.invoke(StandardWrapperValve.java:263)
at org.apache.catalina.core.StandardPipeline.invoke(StandardPipeline.java:551)
at org.apache.catalina.core.StandardContextValve.invokeInternal(StandardContextValve.java:225)
at org.apache.catalina.core.StandardContextValve.invoke(StandardContextValve.java:173)
at org.apache.catalina.core.StandardPipeline.invoke(StandardPipeline.java:551)
at org.apache.catalina.core.StandardHostValve.invoke(StandardHostValve.java:170)
at org.apache.catalina.core.StandardPipeline.invoke(StandardPipeline.java:551)
at org.apache.catalina.core.ContainerBase.invoke(ContainerBase.java:933)
at com.sun.enterprise.web.connector.httpservice.HttpServiceProcessor.process(HttpServiceProcessor.java:234)
at com.sun.enterprise.web.HttpServiceWebContainer.service(HttpServiceWebContainer.java:2124)

Oracle Applications 11i Load Balancing does not work with RAC one Node

Hi all,
Could you help me to resolve this issue.
Architecture environment is :
- One APPS tier node
- Two nodes Oracle Database Appliance (Primary node 1 holds INSTANCE_1 et Secondary node is configurured to holds INSTANCE_2), i.e RAC one Node.
- The primary node have instance_name SIGM_1 and the secondary node have instance_name SIGM_2, but in RAC one node, the secondary instance is not alive.
We convert our EBS 11i environment to RAC following note ID Using Oracle 11g Release 2 Real Application Clusters with Oracle E-Business Suite Release 11i [ID 823586.1].
When testing Database failover, Oracle Applications 11i load balancing does not work anymore.
The root cause is that, when the primary node of the Rac one node is down, the INSTANCE_NAME_1 is automaically relocating to the surviving node,.
During test failover, we imagine that when the primary node goes down, the secondary node start or relocate database with instance_name SIGM_2, and in that case the Oracle Applications load balancing works.
Currently, when the primary node goes down, the instance_name SIGM_1 is relocated on the secondary node, which cause failure of Oracle Applications Load Balancing.
Thank you for your advice.
Moussa

This is something I observed a long time ago for Safari (ie: around version 1). I'm not sure this is Safari, per se, but OpenSSL that is responsible for the behavior. I'm pretty sure Chrome does this and I've seen some Linux browsers do it.
What I have done at the last two companies I've worked for is recommend that our clients do not use SSL SessionID as the way of tracking sticky sessions on web servers, but instead using IP address. This works in nearly all cases and has few downsides. The other solution is to use some sort of session sharing on your web servers to mitigate the issue (which also means that your web servers aren't a point of failure for your users' sessions). (One of the products I supported had no session information stored on the web servers, so we could safely round-robin requests, the other product could be implemented with a Session State Server... but in most cases we just used IP address to load balance with). The other solution is to configure your load balancer to terminate the SSL tunnel. You get some other benefits from this, such as allowing your load balancer to reduce the number of actual connections to the web servers. I've seen many devices setup this way.
One thing to consider through this is that - due to the way internet standards work - this really can't be termed a bug on anyone's part. There is no guarantee in the SSL/TLS standards that a client will return the same SSL Session ID for each request and there is not requirement that subsequent requests will even use the same tunnel. Remember, HTTP is a stateless protocol. Each request is considered a new request by the web server and everything else is just trickery to try and get it to work the way you want. You can be annoyed at Safari's behavior, but it's been this way for over 5 years by my count, so I don't expect it to change.

How Load Balance works in RAC?

Similar Messages

Maybe you are looking for