Http session load balancing

          Hi,
          My application is clustered on multiple machines. The application receives
          request mainly through the http protocol, and the application keeps session information
          of the user. I am using HttpClusterServlet to proxy request to the cluster.
          How can I get the servlet to proxy requests to the right machine that has the
          session information. (Currently, it's just using the round robin algorithm)
          Thanks

          It should do this automatically - it looks at the WebLogicSession (or jsession)
          cookie to determine where to send the request.
          Mike
          "Mark Liu" <[email protected]> wrote:
          >
          >Hi,
          >
          > My application is clustered on multiple machines. The application
          >receives
          >request mainly through the http protocol, and the application keeps session
          >information
          >of the user. I am using HttpClusterServlet to proxy request to the cluster.
          >
          >How can I get the servlet to proxy requests to the right machine that
          >has the
          >session information. (Currently, it's just using the round robin algorithm)
          >
          >Thanks
          >

Similar Messages

HTTPS with load balancing

Hi guys,
We have a portal system with instance 08, so we typically connect to the portal using port 50800 for HTTP, and 50801 for HTTPS.
We have just created a second server node for this portal (in the config tool).
When we connect to 50800, does this automatically load balance the user to the better server? From some reading on these forums, it seemed to indicate that load balancing will only occur if I connect using port 8109. (where 09 is the instance number for the SCS of our portal)
When connecting to port 8109, we are redirected to port 50800, as I'd expect.
Question 1 - do we need to use 8109 for load balancing, or can we still use 50800?
Question 2 - If we need to use 8109, which is a HTTP port, how can we achieve load balancing with HTTPS. Is there a different port we need to use to have HTTPS with load balancing?
Question 3 - Is the creation of a second server node the best way to accomodate additional users and load on the portal system, or is there a better way to do things?
Thanks,
Michael.

Better late than never.
The load balancing you describe through the message server has its limitation. It redirects you to one of the dialog server hosts which means that any bookmarks will always point directly to a dialog server which may be down at that moment.
Access directly to a dialog server on port 50800 will sort of load balanc on the java server instances that are on that server but not on other servers.
The general recommendation is to setup an external loadbalancer and SAP Web dispatcher is a good match if the load is not very high. SAP webdispatcher will then bind up the cluster address and act as a proxy towards the dialog servers of the portal. The user will therefore only see one address. This will also work for HTTPS.
Regards
Dagfinn

Https through load balancer breaks declarative security

Hello,
My desired setup is for a Jboss cluster serving requests behind a load balancer. Also I intend to use declarative security on the deployed units and have ssl client side authentication.
I need someone to please confirm/deny the following statements:
1) ssl has to be negotiated by the load balancer, whether hardware or software based (apache with mod_proxy/mod_jk).
2) if using apache with mod_jk it is possible to configure it to send the client side authentication details (certificate) in such a way that jboss may enforce declarative authorization as if it had done the authentication itself. This also means that the programatic means to get the authenticated user identity described in the ejb and servlet specs will still work.
3) there is no hardware load balancer that supports the behavior described in 2), which means that with a hardware load balancer it is impossible to use declarative authorization enforcement.
After a whole lot testing and digging up for info, I'm quite desperate to solve this question, so if someone could help me I would be most thankfull.
Nuno

After further research, I think the best course of action will be to create a VLAN for the zone behind the BigIP and then create the corresponding interface in the vlan and zone. Using this links as my references in case anyone is interested. I'll post what I come up with.
https://blogs.oracle.com/stw/entry/using_ip_instances_with_vlans
https://blogs.oracle.com/stw/entry/solaris_zones_and_networking_common
http://docs.oracle.com/cd/E19253-01/816-4554/816-4554.pdf # AdministeringVirtualLocalAreaNetworks
http://docs.oracle.com/cd/E19053-01/ldoms.mgr11/820-4913-10/820-4913-10.pdf # Assign VLANs to a Virtual Switch and Virtual
Network Device

Load balancing effect on ssh/https connections

We have a RV016 load balancing between two broadband WAN connections. On protocols that are sensitive to a change in IP address such as ssh and https, if the client connection goes inactive for a short time (sometimes as short as 10 seconds), the RV016 often changes WAN connection as part of its "load balancing" feature. Most protocols do not even notice, but the more sensitive protocols do and often lock a session or timeout the session which is not a good thing.
We have been able to bind these sensitive protcolols to a particular WAN port but (in our minds) this is not an "ideal" situation. In fact I would consider this to be a broken "load balancing" solution and should be fixed.
Does anyone have any "permanent fixes" or ideas on this?
Thanks!

Sorry Tom, but I wrote (edited for clarity): "What would happen [..] when the particular WAN chosen goes down? Does the system apply the same binding to ***another WAN*** until the original one comes back up, will it ignore the bining and so break any new sessions of the same protocol or will it simply fail (no connection)?"
I did not write "the other WAN". That is, I am considering the case, quite common, of more than two WANs.
Proper load balancing is described here:
http://help.mysonicwall.com/sw/eng/305/ui2/23100/Network/WAN_Failover_Load_Balancing.htm
"About Source and Destination IP Address Binding
When you establish a connection with a WAN, you can create multiple interfaces, dividing up the task load over these interfaces. There are both Primary and Secondary WAN interfaces. This task distribution model maintains high performance, ensuring that one interface does not become an impasse to the point where it blocks traffic from passing. This process is WAN Load Balancing.
While WAN Load Balancing addresses performance challenges, it can create other problems, including losing track of sessions. Session confusion can occur because some applications fail to adequately track multiple user sessions load-balanced on multiple interfaces. These applications treat incoming packets as originating from different users because they use IP addresses to differentiate user sessions instead of application-layer user identification tags.
To ensure that you have proper connectivity in all applications, SonicWALL provides a feature called Source and Destination IP Addresses Binding, a solution that maintains a consistent mapping of traffic flows with a single outbound WAN interface."
and their appliances are no more expensive than Cisco multi WAN ones...

Load balance traffic to a service based on a added field in the HTTP Header

I am trying to use HTTP Header Load balancing but the field we want to use in order to load balance is "user-defined", example HTTP_TOTO = toto1.
Do you have any idea on how I could perform this ?
Thanks in advance

Load balancing using pre-defined headers is supported. Not sure if load balancing using user defined fields is possible. You could refer to the following document.
http://www.cisco.com/univercd/cc/td/doc/product/webscale/css/css_710/bsccfggd/httphead.htm
We would appreciate it if someone could share their experience if they know more about this.

11i load balancing web nodes without use of Hardware http load balancer

I am looking at note 217368.1 (Advanced Configurations and Topologies for Enterprise Deployments of E-Business Suite 11i) and some other notes on load balancing but some aspects are not clear.
Aim is to implement load balancing traffic to web nodes without using Hardware ( BigIP, cisco etc) for HTTP layer load balancing.
Which is more preferable between dns or Apache Jserv load balancer ?
Need details like failover capabilities, death detection of node, functionality testing and ways to monitor Apache Jserv load balancer.
Any help in this regard is welcome .
thx
arun

Oracle recommends using loadbalancing hardware rather than using DNS. If you want the features you mention above, you will need a hardware loadbalancer.
http://blogs.oracle.com/stevenChan/2006/06/indepth_loadbalancing_ebusines.html
http://blogs.oracle.com/stevenChan/2009/01/using_cisco_ace_series_hardware_load-balancers_ebs12.html
HTH
Srini

Oracle RAC Concept - load balancing

Hello All,
I have a question about Oracle RAC and how it balance loading.
What I know about Oracle RAC that it balancing load between its node, so if a new session is connected to the Database it will execute on the node which have less load.
If I am having an application that connects to the database with few number of sessions and some sessions have huge (huge load) processes to execute while the other sessions are executing small process (in terms of load).
So below is an example scenario that i am asking for an answer of it:
if am having a two node RAC database.
My application that is connected to this RAC database wants to execute three processes P1, P2 and P3 each will have a session correspondingly S1, S2 and S3
Let is say that P1 will take 60% of the database resources (memory ....) and P2 will take 1% and P3 1%.
So my question is it a sessions load balancing or Processes load balancing ? So in case and randomly (Managed by ORACLE RAC) S1 is connected to node 1 and it is using node 1 resources so the load on node 1 will be 60%, so when S2 and S3 needs to connect it will go to node 2 since node 1 is loaded. in that case i will have Node 1 using 60% of it is resources (becuase of S1 that is executing P1) while node 2 only uses 2 % (because of S2 and S3 that is executing P2 and P3).
Is that how Oracle RAC works ? or it balance the load of S1 (60%) between node 1 and node 2 ?
I am asking this question since my application is not users application, so it will connect to the database with few number of sessions, and 1 session of them may execute a huge process while the other are executing small processes.
So in that case, how RAC is doing the balance loading?
Regards,

i hope , this links may help you.
Thread: Server side Load balancing in RAC
Server Side Load Balancing Testing
read Oracle documentation:
http://www.oracle.com/pls/db102/search?word=server+side+load+balancing&partno=
http://www.oracleracexpert.com/2010/01/oracle-rac-load-balancing-and-failover.html
http://www.databasejournal.com/features/oracle/article.php/3659411/Oracle-RAC-Administration---Part-15-Connection-Load-Balancing-and-FAN.htm
http://oracleinstance.blogspot.com/2010/08/transparent-application-failover-taf.html

MPLS TE equal or unequal load balancing doesn't work? - step2

Previous question in thread:
Dear Sir!
I've two MPLS TE tunnels from one PE to another PE.
And there are traffic share count between them
(as tunnel mpls traffic-eng load-share command define).
But in real life all traffic from the same source to the same destination go through only one tunnel
(as CEF define - i.e. how sh ip cef exact-route says).
PEs are 3660 platforms with c3660-jk9o3s-mz.123-8.T
installed.
How can I correct this problem?
But this answer does not solved my issue:
hritter - Network Consulting Engineer, CISCO SYSTEMS, CCIE
Aug 4, 2004, 7:20am PST
This is expected behavior since CEF is used at the head end to perform label imposition. I wouldn't recommend changing the default bahavior to per=packet loadsharing since this could lead to of of sequence packets, which could lower the overall performance.
Hope this helps,
so my secound question:
Dear Sir!
I'm agree with you as MPLS TE tunnels are opened from PE to PE, so CEF does it work.
But if I open this tunnels from P to PE, ONLY ONE of this tunnels are used instead of load-sharing, if traffic go from one source (of site1 of VPN1) to the same destination (located at site2 of VPN1).
Why? Packet through P-devices swithes by labels, so I mean that CEF cannot does src-dst load sharing?
My problem are that I must to do load sharing between this two tunnels in the case above.
Q: How can I solve this problem?
Best regards,
Maxim Denisov

The per session load-balancing is also used by MPLS when multiple paths are available. Changing this behavior to per-packet is still not recommended.
Hope this helps,

Load Balance Problem in Oracle RAC 10.2.0.4 (two nodes with CRS)

Hi all,
I have an Oracle RAC 10.2.0.4 with two nodes and I'm having problems with sessions load balance.
The problem is that all connections are going to the node which is enabled as 'master' in RAC, the node who has located .db resource in CRS.
If .db resource is located in node1 all connections are made in node1. In the same way, all connections are made in node2 when .db resource from CRS stack is located in node2.
The connections are made by a pool in Tomcat 6 using RAC service for made the connections.
Any idea?
Thanks in advance.
Here is an example of my net configuration:
listener.ora in node1:
LISTENER_RAC_NODE1 =
(DESCRIPTION_LIST =
(DESCRIPTION =
(ADDRESS = (PROTOCOL = TCP)(HOST = node1_vip)(PORT = 1525)(IP = FIRST))
(ADDRESS = (PROTOCOL = TCP)(HOST = node1)(PORT = 1525)(IP = FIRST))
tnsnames in node1:
LISTENERS_RAC =
(ADDRESS_LIST =
(ADDRESS = (PROTOCOL = TCP)(HOST = node1_vip)(PORT = 1525))
(ADDRESS = (PROTOCOL = TCP)(HOST = node2_vip)(PORT = 1525))
NODE1 =
(DESCRIPTION =
(ADDRESS = (PROTOCOL = TCP)(HOST = node1_vip)(PORT = 1525))
(ADDRESS = (PROTOCOL = TCP)(HOST = node1)(PORT = 1525))
(CONNECT_DATA =
(SERVER = DEDICATED)
(SERVICE_NAME = rac)
(INSTANCE_NAME = rac1)
RAC =
(DESCRIPTION =
(ADDRESS = (PROTOCOL = TCP)(HOST = node1_vip)(PORT = 1525))
(ADDRESS = (PROTOCOL = TCP)(HOST = node2_vip)(PORT = 1525))
(LOAD_BALANCE = yes)
(CONNECT_DATA =
(SERVER = DEDICATED)
(SERVICE_NAME = rac)
********************************************************

It is my fault.
It was the connection service. I had one one with preferred option and one without it. Now is working well.

Hardware load balancing with session stickiness

Hello,
          We are looking for a hardware based solution to load balance 4
          unclustered weblogic servers. These servers are software replicas of
          each other and share a common database. We need a solution that
          provides session level server stickiness based on jsessionid as both a
          cookie and url rewrite. Currently we have Cisco load balancers which
          do not handle the url rewrite due to the fact the the load balancer
          only "sees" what is after the question mark in the url and the
          jsessionid is before the question mark. Example:
          http://somehost.com/some/url.jsp;jsessionid=26D5C566075663ABD8E17AD255974323?parm=value
          I'm sure we are not the only ones trying to do this. Any suggestions
          would be greatly appreciated.
          Thanks,
          Alan George


          Having worked on Cisco load balancer and WLS clustered, you would loose sessions
          when the requests are failed over.
          S
          "Shawn Kircher" <[email protected]> wrote:
          >
          >Has anyone got this type of scenario working where you load balance sticky
          >sessions
          >from a Cisco load balancer between non-clustered weblogic servers?
          >
          >Thanks,
          >
          >Shawn
          >
          >"Aravind Krishnasamy" <[email protected]> wrote:
          >>F5 or BigIp provides hardware level load balancing and it injects date
          >>inside weblogic cookie and maintains server affinity.
          >>Thanks
          >>Aravind
          >>"ageorge" <[email protected]> wrote in message
          >>news:[email protected]...
          >>> Hello,
          >>>
          >>> We are looking for a hardware based solution to load balance 4
          >>> unclustered weblogic servers. These servers are software replicas
          >>of
          >>> each other and share a common database. We need a solution that
          >>> provides session level server stickiness based on jsessionid as both
          >>a
          >>> cookie and url rewrite. Currently we have Cisco load balancers which
          >>> do not handle the url rewrite due to the fact the the load balancer
          >>> only "sees" what is after the question mark in the url and the
          >>> jsessionid is before the question mark. Example:
          >>>
          >>>
          >>http://somehost.com/some/url.jsp;jsessionid=26D5C566075663ABD8E17AD255974323
          >>?parm=value
          >>>
          >>> I'm sure we are not the only ones trying to do this. Any suggestions
          >>> would be greatly appreciated.
          >>>
          >>> Thanks,
          >>>
          >>> Alan George
          >>
          >>
          >

Load balancing and session sharing

Hi,
in my company we have the following situation:
We have 2 instances of the same web application on differend servers.
An external load balancer (Alteon) routes the requests to these instances.
The problem:
If one instance is shutting down, all the sessions of this app are gone. These users has to login again.
The question:
Is it possible to avoid this?
Is is possible to keep these session, so the users don't have to login again?
Maybe with a kind of session sharing between these 2 instances?
using:
oc4j standalone - 10.1.3.1
OS: Windows
thank you
Marcus

The solution:
Multicast IP replication is the default replication protocol used in a standalone OC4J installation
see: http://download.oracle.com/docs/cd/B32110_01/web.1013/b28950/cluster.htm#BABBHGHJ
-> Configuring Multicast Replication

Load Balancing 2012 R2 Session Host Collection with External Network Load Balancer

Hi,
We are moving from a 2008 R2 Remote Desktop session host deployment to 2012 R2. Previously, we used our Kemp hardware load balancer to distribute load between RDSH servers. We had a connection broker deployed so that if an existing disconnected session was
detected during the initial connection, the user was directed back to that session.
In 2012 R2, we planned to again used the Kemp load balancer to main high availability for our RDSH collection, but are experiencing strange issues. It seems that the RD Connection Broker is also performing load balancing--the result being that initial connections
to the RDSH collection may go to one RDSH server with the least connections through the Kemp, but then be redirected to a different RDSH server by the broker, even when there is no existing session for the user on that second server.
Our question is: Should we not be using the Kemp balancer at all (how would this work)? Or should we disable load balancing by the connection broker (if so...how)?
Further complicating our redirection issue with that the RDSH servers have multiple interfaces--one with public addresses and others with private. The connection broker seems to abritrarily pick among the destination RDSH server's available IP addresses
for the redirection and trying to redirect to a private address will fail. We think we have worked around this by connecting to each RDSH server from a 2008 R2 server's RDSH Configuration console and choosing just the public adapter under the Network Adapters
tab--is there no way to access this setting in 2012 R2?
Thanks in advance!
Matthew

Hi Matthew,
As you are most likely already aware, inn Remote Desktop Services 2012 / R2 the Connection broker uses round robin DNS to load balance.
To simplify things I would recommend that you let the connection broker load balance the sessions and use the KEMP to Load balance the RDweb and Gateway servers.
Have a look at the following articles:
http://ryanmangansitblog.wordpress.com/2013/03/11/create-a-rdwa-farm-using-a-kemp-load-balancer/
http://ryanmangansitblog.wordpress.com/2013/03/31/rds-2012-configuring-a-rd-gateway-farm/
http://ryanmangansitblog.wordpress.com/2013/09/05/load-balance-rds2012-rdwa-and-rdgw-using-sub-interfaces-on-kemps-loadmaster/
As you have mentioned that you are migrating from a 2008R2 configuration, have a look at the following article:
http://ryanmangansitblog.wordpress.com/2014/01/05/publish-rds-2008r2-desktop-on-rds-2012/
Ryan Mangan | [email protected] | Help keep the forums tidy, if this has helped please mark it as an answer

Sticky sessions and Load Balancing in WL Clusters

We are using iPlanet Web Server 4.1 with WebLogic App Server; and would like
to implement load balancing with sticky sessions and in-memory state
replication.
The documentation in Weblogic says that -
When using in-memory state replication, your WebLogic Server Cluster must
live behind one or more proxy servers. The proxy servers are smart enough to
send servlet requests, belonging to the same HTTP session, back to the same
server in the cluster that holds the session data.
(Ref: http://www.weblogic.com/docs51/cluster/setup.html)
Does this mean that the sticky session configuration has to be done on the
iPlanet Web Server itself ?
Also, if WebLogic is used as the Web server, does WebLogic provide any
support for sticky sessions?
Any help, suggestions or links to useful info are welcome.
Regards,
Milind.

Mike,
im curious as to why you would recomend using weblogic as a web server in 6.1?
I would not for the following reasons:
- it costs 10x more per cpu list
- it doesnt support hardware accell cards (afaik, please let me know if this has
changed)
iplanet is really good a serving up static html and gif's, especially in ssl if you
have a hardware accell card. So if you have a site with lots of graphics and you use
ssl a lot, I think its still a better solution.
-Joel
Mike Reiche wrote:
You get sticky round-robin by default.
You need to have session tracking turned on (i think it is on by default). You
need to have the WL plugin configured in iPlanet.
When WL creates an httpSession, it writes a cookie (or rewrites the URL) back
to the browser. On subsequent requests, the browser sends the cookie and iPlanet
plug-in directs the request to the correct WL instance based on the ip address
of the WL server embedded in the cookie.
If you are using WLS 6.1, I would recommend using it as a web server (and not
using iPlanet). I imagine that it supports stickly load balancing as well.
Mike
Joel Nylund <[email protected]> wrote:
you get round robin by default, if you want a different scheme you can
use one
of the other 3 options (weight, random or parameter).
-Joel
I think weight can be set in weblogic properties. I havent used any other
than
round robin.
Milind Prabhu wrote:
We are using iPlanet Web Server 4.1 with WebLogic App Server; and wouldlike
to implement load balancing with sticky sessions and in-memory state
replication.
The documentation in Weblogic says that -
When using in-memory state replication, your WebLogic Server Clustermust
live behind one or more proxy servers. The proxy servers are smartenough to
send servlet requests, belonging to the same HTTP session, back tothe same
server in the cluster that holds the session data.
(Ref: http://www.weblogic.com/docs51/cluster/setup.html)
Does this mean that the sticky session configuration has to be doneon the
iPlanet Web Server itself ?
Also, if WebLogic is used as the Web server, does WebLogic provideany
support for sticky sessions?
Any help, suggestions or links to useful info are welcome.
Regards,
Milind.

Load balancing http server

Hi There,
I want to implement a load balancer (linux virtual server) into our htmldb configuration. We currently run the http server on the same machine running oracle/htmldb. I want to split this out and use 2 seperate machines running http server behind the load balancer, both these http servers will be pointing to the same oracle database.
The load balancer will not be using persistent sessions, therefore client requests will be sent evenly to each http server. My main question is, will I have problems with user sessions, or (as I imagine) is all the session information written to the database ?
Thanks in advance.
Tom

Bill,
There are a few things you need to consider
1: Availability of http server. ie how many are you going to have. How are you going to access them. Load balancer, DNS round robin. These should be on different servers to the database.
2: How are the http servers going to connect. For RAC you'll need to specify TNS connections. I'd also recommend that you look at using application partitioning using services in the RAC cluster. That way you can have Apex using a subset of nodes in the cluster .
3: If you are going down the RAC path then I'd assume availability is a priority. You''ll need to think of standby configuration. Again this is possible with TNS configuration.
You can do it all with Apex. RAC and Standby but it will take planning and testing.

Apache HTTP proxying for load balancing only to a group of non-clustered WL servers

          Hi,
          We're running WL Server 6.1 SP 2 on Solaris 2.8.
          For the Apache HTTP proxy plugin, if you use the WebLogicCluster http.conf option,
          do the WL servers you want to load balance across have to be part of a WebLogic
          cluster (if you are prepared to do without failover, as I know it would need to be
          a proper WL cluster to replicate session info for failover). Can you load balance
          across a group of non-clustered WL servers, and maintain the user session to the
          one WL server so that it doesn't switch between servers on alternate requests for
          the same user session, or must the servers be configured as a WebLogic cluster?
          Paul
          We find that if you have a collection of WL servers that are not configured as a
          cluster, that it will load balance alternate requests to each server, but it will
          not pin a user to a single machine according to their session so for 2 servers, 2
          differetn sessions get created, one on each machine.
          Is this because it doesn't normally do this, but sends the user alternately to a
          primary then secondary which works in a cluster because the session is replicated.
          I thought the secondary was only used when the primary failed.


We're running WL Server 6.1 SP 2 on Solaris 2.8.          >
          > For the Apache HTTP proxy plugin, if you use the WebLogicCluster http.conf
          option,
          > do the WL servers you want to load balance across have to be part of a
          WebLogic
          > cluster (if you are prepared to do without failover, as I know it would
          need to be
          > a proper WL cluster to replicate session info for failover). Can you load
          balance
          > across a group of non-clustered WL servers, and maintain the user session
          to the
          > one WL server so that it doesn't switch between servers on alternate
          requests for
          > the same user session, or must the servers be configured as a WebLogic
          cluster?
          You don't have to use the clustering option. To get failover, you'll have to
          use the JDBC persistence option of WL.
          > We find that if you have a collection of WL servers that are not
          configured as a
          > cluster, that it will load balance alternate requests to each server, but
          it will
          > not pin a user to a single machine according to their session so for 2
          servers, 2
          > differetn sessions get created, one on each machine.
          >
          > Is this because it doesn't normally do this, but sends the user
          alternately to a
          > primary then secondary which works in a cluster because the session is
          replicated.
          > I thought the secondary was only used when the primary failed.
          The primary/secondary stuff requires clustering. If Apache continues to
          "load balance" after the first request, you need to either use JDBC session
          persistence or use a different load balancer (like mod_jk for Apache or a
          h/w load balancer with support for sticky).
          Peace,
          Cameron Purdy
          Tangosol, Inc.
          http://www.tangosol.com/coherence.jsp
          Tangosol Coherence: Clustered Replicated Cache for Weblogic
          "Paul Hammond" <[email protected]> wrote in message
          news:[email protected]...
          >

Http session load balancing

Similar Messages

Maybe you are looking for