WLS Cluster Servlet not "sticky" - continuous round robin (WLS 6.1 SP2)

          We have WLS serving as a proxy (with HttpClusterServlet) to a cluster of WLS servers
          running a WebApp with in memory HttpSession replication.
          However, subsequent calls from a browser continue to round robin through the cluster
          instead of returning to the same server.
          We have enable cookies in the browser.
          We have configured replication in the web.xml file.
          Is this an issue with SP2 perhaps?
          thanks in advance,
          -Mark

          We've noticed this as well, see my recent post. It definetly seems to be different
          from WL5.1.
          Sometimes we also see the same HTTPRequest going to BOTH servers in our cluster.
          We were testing with only one browser window and one user, and occasionally both
          clusters process the same request.
          We are using IIS with the proxy plugin and Round Robin. We installed the WL6.1
          verison of the plug in aswell.
          It would be great if you could configure WL6.1 to be sticky where one user sticks
          to one server until failover.
          "Mark" <[email protected]> wrote:
          >
          >We have WLS serving as a proxy (with HttpClusterServlet) to a cluster
          >of WLS servers
          >running a WebApp with in memory HttpSession replication.
          >
          >However, subsequent calls from a browser continue to round robin through
          >the cluster
          >instead of returning to the same server.
          >
          >We have enable cookies in the browser.
          >
          >We have configured replication in the web.xml file.
          >
          >Is this an issue with SP2 perhaps?
          >
          >thanks in advance,
          >-Mark

Similar Messages

Http cluster servlet not failing over when no answer received from server

          I am using weblogic 510 sp9. I have a weblogic server proxying all requests to
          a weblogic cluster using the httpclusterservlet.
          When I kill the weblogic process servicing my request, I see the next request
          get failed over to the secondary server and all my session information has been
          replicated. In short I see the behavior I expect.
          r.troon
          However, when I either disconnect the primary server from the network or just
          switch this server off, I just get a message back
          to the browser - "unable to connect to servers".
          I don't really understand why the behaviour should be different . I would expect
          both to failover in the same manner. Does the cluster servlet only handle tcp
          reset failures?
          Has anybody else experience this or have any ideas.
          Thanks


I think I might have found the answer......
The AD objects for the clusters had been moved from the Computers OU into a newly created OU. I'm suspecting that the cluster node computer objects didn't have perms to the cluster object within that OU and that was causing the issue. I know I've seen cluster
object issues before when moving to a new OU.
All has started working again for the moment so I now just need to investigate what permissions I need on the new OU so that I can move the cluster object in.

Round Robin doesn't work in Bea Weblogic Cluster

hello!
          we are using a weblogic (V5.1 SP8) cluster in the following configuration:
          2 Weblogic App-Servers as Servlet Containers.
          4 Weblogic App-Servers as EJB Containers.
          if a client connects to the first servlet - server, round robin between the
          4 EJB-servers behind doesn't work. It seems that the servlet-server is
          pinned to one EJB-server all the time.
          If we kill the second servlet-server as from now on round robin works in the
          EJB-server cluster like we expected.
          As we changed this configuration and also use an IPlanet webserver (one
          instance) with the weblogic proxy-plugin which connects to the
          servlet-cluster (2 instances)... round robin is working properly in the
          EJB-server cluster when the client connects to the IPlanet-server....
          what is the problem, why doesn't round robin work in the first
          configuration? any suggestions?
          Torsten Jenkner


Okay, is it that the round-robining is not perfect (i.e., it doesn't send the
          first request to server1 and the second request to server2 and so on) or is it
          sending every request to one server?
          Torsten Jenkner wrote:
          > yes exactly Robert that's it!
          >
          > Torsten
          >
          > "Robert Patrick" <[email protected]> schrieb im Newsbeitrag
          > news:[email protected]...
          > > I don't understand this. Are you saying that with servlet node 1 or
          > servlet
          > > node 2 running by themselves that RR works fine but once you start both
          > > together, it does not?
          > >
          > > This sounds like a case for support...
          > > Robert
          > >
          > > Torsten Jenkner wrote:
          > >
          > > > > > it is always the same stub in our case. Also the round robin in the
          > EJB
          > > > > > cluster does with the same java classes but does not if the servlet
          > > > cluster
          > > > > > consits of two nodes. So it seems to be a configuration problem, but
          > > > which
          > > > > > one?
          > > > >
          > > > > Hmm.... I don't quite understand this last statement about the EJB
          > and
          > > > servlet
          > > > > cluster, could you explain a little more?
          > > > >
          > > > > Thanks,
          > > > > Robert
          > > > >
          > > >
          > > > if you use 2 servlet cluster nodes round robin between the
          > servlet-cluster
          > > > and the ejb-cluster doesn't work. if i kill the second servlet-node it
          > works
          > > > fine. so my thought was it has to be a configuration problem, because RR
          > > > does with one servlet-node. ok?
          > > >
          > > > Torsten Jenkner
          > >

Round Robin was not happening for my cluster with WebLogic Proxy Plugin

          Hi,
          I configured my cluster with software load balancer, HTTPClusterServlet. By default,
          it is load balancing with Round Robin Algorithm. That means one HTTP request
          goes to server1, and the other HTTP request goes to server2. However, it is not
          what I can see no matter for the requests in one HTTP session or not.
          Say I open two browser, and log into my application with two different users,
          one is "cyang", the other is "xpression". Then the HTTP request (for servlet/jsp)
          from two browsers always go to the same server1, server2 is not invoked at all.
          I did see one time, with only one session (one browser with "xPression" user log
          in), the most requests go to server1, suddenly I am brough into log in page, then
          I noticed that the request for "xPression" moved to server2 although server1 is
          still alive. Therefore, at most, I can say it is "Random", rather than "Round
          Robin".
          What is the real meaning for HTTP servlet/jsp load balancing algorithm? Does
          Round Robin mean request go to each server in turns? Does it have to be different
          session or it can be within one session?


It should be sticky. If not, then bug / config error.
          Peace,
          Cameron Purdy
          Tangosol, Inc.
          http://www.tangosol.com/coherence.jsp
          Tangosol Coherence: Clustered Replicated Cache for Weblogic
          "Carole Yang" <[email protected]> wrote in message
          news:[email protected]...
          >
          > Thanks. Yeh, with two kind of browsers, I do see the request goes to
          different
          > servers.
          >
          > That goes back to the original question. Does "Round Robin" here fall
          into the
          > scope of a HTTP session. However, that is not always true for my tests.
          Sometimes,
          > HTTP request just randomly goes to another server while in the same HTTP
          session.
          > It is not sticky to one particular server during one session.
          >
          >
          >
          > --Carole
          >
          >
          > "Cameron Purdy" <[email protected]> wrote:
          > >Has to be different sessions to go to different machines.
          > >
          > >Best way to test is to run one session in IE and the other in Mozilla
          > >or
          > >Netscape.
          > >
          > >Peace,
          > >
          > >Cameron Purdy
          > >Tangosol, Inc.
          > >http://www.tangosol.com/coherence.jsp
          > >Tangosol Coherence: Clustered Replicated Cache for Weblogic
          > >
          > >
          > >"Carole Yang" <[email protected]> wrote in message
          > >news:[email protected]...
          > >>
          > >> Hi,
          > >>
          > >> I configured my cluster with software load balancer,
          HTTPClusterServlet.
          > >By default,
          > >> it is load balancing with Round Robin Algorithm. That means one HTTP
          > >request
          > >> goes to server1, and the other HTTP request goes to server2. However,
          > >it
          > >is not
          > >> what I can see no matter for the requests in one HTTP session or not.
          > >>
          > >> Say I open two browser, and log into my application with two different
          > >users,
          > >> one is "cyang", the other is "xpression". Then the HTTP request (for
          > >servlet/jsp)
          > >> from two browsers always go to the same server1, server2 is not invoked
          > >at
          > >all.
          > >>
          > >> I did see one time, with only one session (one browser with "xPression"
          > >user log
          > >> in), the most requests go to server1, suddenly I am brough into log
          > >in
          > >page, then
          > >> I noticed that the request for "xPression" moved to server2 although
          > >server1 is
          > >> still alive. Therefore, at most, I can say it is "Random", rather
          > >than
          > >"Round
          > >> Robin".
          > >>
          > >> What is the real meaning for HTTP servlet/jsp load balancing algorithm?
          > >Does
          > >> Round Robin mean request go to each server in turns? Does it have
          > >to be
          > >different
          > >> session or it can be within one session?
          > >
          > >
          >

ACE howto: ¿Classic round-robin if cookie not present?

Hello team:
My server farm behind the ACE will be used for applications that may or may not send cookies.
So I have been asked to configure the ACE to be aware of a certain cookie in order to stick client to server as expected. In paralell, I have also been asked to execute a classic round robin algorithm if no cookie is present on the HTTP sessions.
Then my question is: ¿ will my ACE, configured with the "sticky http-cookie ..." command, default to a classical round-robin algorithm if no cookie is detected in the HTTP request from the client?
Your hints will be greatly appreciated.
Best regards, Rogelio

Rogelio,
*/ Create a http parameter-map, if you did not do so already:
parameter-map type http HTTP
case-insensitive
persistence-rebalance
set header-maxparse-length 65535
set content-maxparse-length 65535
length-exceed continue
*/ Match the cookie, and whatever value PHP created:
class-map type http loadbalance match-any PHPSESSID
2 match http cookie PHPSESSID cookie-value ".*"
*/ Create a policy map, add the PHPSESSID class-map, and assign the sticky serverfarm.
*/ Assign your real servers to the default class (class-default)
policy-map type loadbalance http first-match STICKY_SLB_LOGIC
class PHPSESSID
sticky-serverfarm STICKY_SERVERS
class class-default
serverfarm REAL_SERVERS
*/ Add it to the service policy
policy-map multi-match CLIENT_VIPS
class VIP-30
    loadbalance vip inservice
    loadbalance policy STICKY_SLB_LOGIC
    loadbalance vip icmp-reply active
    appl-parameter http advanced-options HTTP
Cheers mate,
Søren Elleby Sørensen

Cluster ( Round Robin with 9ias 9.0.4 )

Hi,
I'm testing clustering with Oracle 9ias 9.0.4.
I have got an oracle 9 ias instance with two oc4j instances.
I have created a cluster and i have added the 9ias instance
to the cluster.
I have added a web application on both oc4j instances
( with <distributable/> in web.xml ).
I've configured the session replication in global-web-application.xml ( on both oc4j instances ):
<cluster-config host="239.0.0.1" port="9129"/>
In web-site.xml of each oc4j instance, i have added :
cluster-island="1" and :
<frontend host="192.168.0.1" port="80"/>
I've configured mod_oc4j like this :
Oc4jSelectMethod roundrobin
Oc4jMount /testcluster
cluster://cluster1:oc4j1,cluster1:oc4j2
Oc4jMount /testcluster/*
cluster://cluster1:oc4j1,cluster1:oc4j2
My problem : all http request goes on the same
oc4j instance. There is no round robin between
the oc4j instances.
If i stop oc4j1, the request goes on oc4j2 and the
session is kept( this is ok ).
But when both instances are running, there is no round
robin at all.
Did i miss something ?
Thanks

Hi Laurent,
Not completely sure of your setup so I'll make some assumptions and then some suggestions.
If you only have one 10g (9.0.4) instance then there's no need to put this in a cluster -- you would have a cluster of 1. I'm assuming the OC4J instances you reference aren't standalone instances but instances that you've created in the 10g instance using either Enterprise Manager or via dcmctl. Have you deployed the same application to both OC4J instances?
If each of your OC4J instances only has one JVM then I would suggest simply creating one OC4J instance ("home" is the default instance) and then configure two JVM's to run under this OC4J instance. This can be done via EM or editing the opmn.xml file (run "dcmctl updateConfig") after doing this.
In this type of configuration you won't have to go in and make the edits you're currently doing in mod_oc4j.conf.
A couple of other things...
Are you using a stateful or stateless app? I'm assuming it's stateful based on your configuring the OC4J config files. If it's stateful then OHS won't load balance requests across JVM's for a session. It detects that a stateful app is deployed and for efficiency routes subsequent requests to the same JVM for efficiency.
Your setup in web.xml is OK.
You should be updating orion-web.xml instead of global-web-application.xml. When you update it all you need to put in is the cluster-config tag. Don't put in the host and port attributes.
You don't need to do anything in http-web-site.xml
Again, I'm not sure if you're trying to work with a mix of OC4J standalones and a 10g instance. Try the above suggestions and also take a look at the "Oracle HTTP Server Admin Guide" and the "Oracle Application Server High Availability Guide" for more information.
John

Isdn bchan-number-order ascending round-robin is applied but not working

isdn bchan-number-order ascending round-robin is applied but not working
ios c3825-spservicesk9-mz.124-21.bin
relevant config
controller E1 0/3/0
framing NO-CRC4
pri-group timeslots 1-31
interface Serial0/3/0:15
no ip address
encapsulation hdlc
no logging event link-status
no snmp trap link-status
isdn switch-type primary-net5
isdn overlap-receiving
isdn incoming-voice voice
isdn bchan-number-order ascending round-robin
trunk-group 1
no cdp enable
my problem is this:-
At the moment calls are leaving on the first available channel of my h323 voice gateway PRI, I can apply this command to the interface fine but it makes no difference the b cannels are always being picked on a first come first served basis . I have also tried combination "isdn bchan-number-order decending" and "isdn bchan-number-order descending round-robin" but it makes no difference , the first b channel is always selected .
Any ideas , if you think its a bug can u identify the bug id as I cannot find it .

Hello
I can see that you use ascending , that why you hit 1st channel.You have to choose one of these commands ascending or descending
I have a question here , for your incoming calls which channel hits ?.FYI isdn bchan-number-order descending is enable by default , so if you need to change you have to type ascending. Can you just share incoming calls which channel hits?.This is because before you define your outgoing calls , you have to check your service provider channel used, and this to avoid error channel is not available.
Command
Router(config-if)# isdn bchan-number-order {ascending | descending}
Purpose
Enables B channel selection for outgoing calls on a PRI interface (optional).
Thanks
please rate all useful information

Command Tilde round robins between windows of an application but does not work in Most Recently Used mode

Hi All,
Suppose I have 5 excel sheets open and I want to toggle between 2 excel sheets on which i'm working on currently. How do it do it in Mac ? Cmd-~ round robins between all the open excels. Alt-Tab in windows work on the Most Recently Used concept. Is there a way to achieve this in Mac ?
Regards,
Avinash

Not a reply to this specific question, just a quasi-relevant comment that might help some folks.
In Leopard, and with (for example) English and Russian chosen in System Prefs under "Language" and the keyboard shortcut set to command+space, I would always see the flag change in the menu bar as soon as command+space were chosen. Now, under Snow Leopard, hitting command+space seems not to make the change. But it does make the change, you just need to take your fingers off the keys first, then the flag will change.

Exchange 2013 CAS - Round Robin DNS not working properly

I have exchange 2013 server (2MB, 2CAS) server. I created two dns records for mail.test.com, autodiscover.test.com pointing to my two CAS servers.
But the problem is if i switched of one cas server, client outlook not connecting automatically to other CAS server. By restarting the outlook also its not working. By restarting the system or running the command ipconfig /flushdns in command prompt, it
working.
is there any configuration iam missing, please advice how to achieve decent load balancing in Exchange 2013 CAS without going for third party Loadbalancer...

I have exchange 2013 server (2MB, 2CAS) server. I created two dns records for mail.test.com, autodiscover.test.com pointing to my two CAS servers.
But the problem is if i switched of one cas server, client outlook not connecting automatically to other CAS server. By restarting the outlook also its not working. By restarting the system or running the command ipconfig /flushdns in command prompt, it
working.
is there any configuration iam missing, please advice how to achieve decent load balancing in Exchange 2013 CAS without going for third party Loadbalancer...
If a CAS role server is down or unable to service clients, you have to remove it from DNS round-robin consideration manually. There is no health check with DNS round-robin unlike a true load balancer.
Also, I would set the TTL to a low value for the CAS servers in the round-robin.
Twitter!: Please Note: My Posts are provided “AS IS” without warranty of any kind, either expressed or implied.

NSAPI not sticky in a cluster

Hi!
I am using NSAPI(IPlanet on SOlaris) as web server as proxy and to serve static contents. My application lies on a cluster with two nodes of managed servers. NSAPI acting as a load-balancer needs to follow the sticky algorothm where it should direct the request from same client to the same node. It should use the session id or cookie to identify which server the session should be sticky to. But NSAPI rotates the requests between the two nodes in random. Since my session is replicated, this doesn't log the user out, but this behaviour is undesirable and not according to the bea docs.
Can anyone suggest how to debug this? Is there anything required on the plug-in side to be configured?
Regards,
Manish Agarwal.

Thanks Eric, but I still have not been able to figure this out. There is something
i want to add here, i checked with the vendor hosting our web serevers and he
replied that there is a hardware loadbalancer in front of the three web servers
distributing the requests between different servers based on load. So now that
the hardware loadbalancer is not sticky based on session id and cookie, can NSAPI
figure out which weblogic node did the last request go to.
Does the hardware loadbalancer need to support sticky sessions?
"Eric Gross" <[email protected]> wrote:
Not possible unless your configurations for the 3 Iplanets are different.
Set Debug=ALL and get the wlproxy.log.
Regards,
Eric
"Manish Agarwal" <[email protected]> wrote in message
news:[email protected]..
I just want to add that i have three instances of NSAPI as Web-Servers,
could it be possible that requests from client going through differentweb
servers instances cannot be sticky? That IPlanet cannot itself be sticky?
Manish Agarwal.

Session not Sticky with HttpClusterServlet

Hi,
I am using HttpClusterServlet for load balancing. The cluster is configured to Round-Robin algorithm. As per my understanding, the weblogic implements the Round-Robin with Sticky session.
But the problem I am facing is that the session is not sticky and lost if the application is accessed through the cluster. (There is no problem if I point directly to the Managed instance / If only one managed instance of the cluster is started)
Is there anything o do with weblogic configuration to have sticky session?
Please give me some clue, pointers to move ahead. Thanks is advance.
Regards,
Sudhakar P

Hi All,
By default HttpClusterServlet will manage the sticky session by "HttpSession" object. If you are getting any problem with that, we probably need to debug more.
1) Can you give me some setup details how you have set it up ..
2) Also if possible, use Fiddler and get me the fiddler output.
3) Why do you think that it is not doing sticky session,are you getting flipped to another node and session data is not avaible.
Also some traces of logs when it happen...
Thanks

Cluster servlet only sees one server

          I'm using WLS6.1 SP2 running on Solaris. I'm using the new version of the HttpClusterServlet
          (weblogic.servlet.proxy.HttpClusterServlet) to proxy to 6 managed servers running
          on two physical servers. It's just proxying web requests, not EJBs or anything
          like that, and we're not using session state or SSL.
          The problem is that the cluster servlet only sends requests to one server at a
          time - no round-robin is going on. It seems to just latch onto the first server
          it can find. So if I bring down all the managed servers, the first one that comes
          back up gets all the requests forever.
          The only parameters to the HttpClusterServlet are DebugConfigInfo = On and the
          WebLogicCluster property which is set to:
          server1:9221:9222|server1:9231:9232|server1:9241:9242|server2:9221:9222|server2:9231:9232|server2:9241:9242
          I've also checked the online debug info, using http://server1:proxyport/dummy.jsp?__WebLogicBridgeConfig.
          This shows only one server under the General Server List. However it does change
          if I take down the server listed there.
          Any help would be greatly appreciated,
          BKR


Hi Brian,
          One thing I notice is that the servers listed in your WebLogicCluster
          property show some instances listening on different ports. Pre-7.0,
          clustered instances must be listening on different IPs, but the same port
          number. For example,
          server1 - listen address 192.168.1.100 - listening on port 8001
          server2 - listen address 192.168.1.101 - listening on port 8001
          server3 - listen address 192.168.1.102 - listening on port 8001
          ..etc...
          So you are either trying to proxy to three different clusters, in which
          case, your HttpClusterServlet configuration is incorrect, or you just have
          your cluster configured in correctly (if you intend for all six servers to
          be a part of one cluster).
          Hope this helps.
          Jane
          BEA Support
          "Brian Reischl" <[email protected]> wrote in message
          news:[email protected]...
          >
          > I'm using WLS6.1 SP2 running on Solaris. I'm using the new version of the
          HttpClusterServlet
          > (weblogic.servlet.proxy.HttpClusterServlet) to proxy to 6 managed servers
          running
          > on two physical servers. It's just proxying web requests, not EJBs or
          anything
          > like that, and we're not using session state or SSL.
          >
          > The problem is that the cluster servlet only sends requests to one server
          at a
          > time - no round-robin is going on. It seems to just latch onto the first
          server
          > it can find. So if I bring down all the managed servers, the first one
          that comes
          > back up gets all the requests forever.
          >
          > The only parameters to the HttpClusterServlet are DebugConfigInfo = On and
          the
          > WebLogicCluster property which is set to:
          >
          server1:9221:9222|server1:9231:9232|server1:9241:9242|server2:9221:9222|serv
          er2:9231:9232|server2:9241:9242
          >
          >
          > I've also checked the online debug info, using
          http://server1:proxyport/dummy.jsp?__WebLogicBridgeConfig.
          > This shows only one server under the General Server List. However it does
          change
          > if I take down the server listed there.
          >
          >
          > Any help would be greatly appreciated,
          > BKR

WebLogic Apache Round Robin Failure

I am using a cluster configuration of WebLogic.
I've been testing the fail-over of weblogic in two cases:
- The two members of the cluster are running and answering to requests
in
round-robin without any problem. One client is doing a request.
Monitoring
the weblogic.log files, we stop (kill) one application server (the one
the
client was talking to).
=> Then the next requests are sent to the other member. OK
- The two members of the cluster are running and answering to requests
in
round-robin without any problem. One client is doing a request.
Monitoring
the weblogic.log files, we stop (disconnect from the network) one
application server (the one the client was talking to).
=> The next request goes to the previous app server (white page for a
really long time / time out).
If another client is doing a request (different IP address) it will go
to the app server still
connected -> OK.
Next client will however go to the app server disconnected (idem,
white
page...).
And so on, OK, time out, OK, time out.
Therefore it seems that our bridge is still doing round-robin on the two
app
server, and that it didn't notice that one was not on the network
anymore if
the server physically disappears (instead of the first case were the
server
is still there but nothing listening on port 7001 where it works). We
didn't find any particular setting that allow to control the behaviour
of
the bridge (e.g. timeout parameters...) and I will be glad to learn more
on
the possible tuning of it.
Moreover, do you have any idea about how the application server is
talking to the web server. I have heard that the web server was using
cookies sent by the application server to check its availabilities.
Therefore, how these cookies are refreshed?
Thank you for your help.
Erwan,

I am using a cluster configuration of WebLogic.
I've been testing the fail-over of weblogic in two cases:
- The two members of the cluster are running and answering to requests
in
round-robin without any problem. One client is doing a request.
Monitoring
the weblogic.log files, we stop (kill) one application server (the one
the
client was talking to).
=> Then the next requests are sent to the other member. OK
- The two members of the cluster are running and answering to requests
in
round-robin without any problem. One client is doing a request.
Monitoring
the weblogic.log files, we stop (disconnect from the network) one
application server (the one the client was talking to).
=> The next request goes to the previous app server (white page for a
really long time / time out).
If another client is doing a request (different IP address) it will go
to the app server still
connected -> OK.
Next client will however go to the app server disconnected (idem,
white
page...).
And so on, OK, time out, OK, time out.
Therefore it seems that our bridge is still doing round-robin on the two
app
server, and that it didn't notice that one was not on the network
anymore if
the server physically disappears (instead of the first case were the
server
is still there but nothing listening on port 7001 where it works). We
didn't find any particular setting that allow to control the behaviour
of
the bridge (e.g. timeout parameters...) and I will be glad to learn more
on
the possible tuning of it.
Moreover, do you have any idea about how the application server is
talking to the web server. I have heard that the web server was using
cookies sent by the application server to check its availabilities.
Therefore, how these cookies are refreshed?
Thank you for your help.
Erwan,

RD Web Access / RD Connection Broker - NLB/Round Robin or both?

Hi
Can someone point me to documentation that can assist with Infrastructure scaling and HA config of the RD Web Access/RD CB roles?
We have a successful PoC consisting of RD Web and RD CB on a single server (no GW required as we're using UAG) and several hosts. Now looking to port to production, thinking of scalability and HA. Looking to serve 300-750 users.
Thinking so far:
RD Web Access role across 2 servers using NLB
RD Connection Broker across 2 servers - (maybe same as Web Access machines?) But CB appears to operate using DNS Round robin. (https://msfreaks.wordpress.com/2013/12/23/windows-2012-r2-remote-desktop-services-part-2/)?
Can CB use NLB?
Which is preferred for the CB role, NLB or DNS round robin?
Cheers
Lea

Thanks Amy, they're really useful links. Why is it so hard to find these....or even an implementation/scaling guide for 2012?
TP - you are my hero, thanks for answering so many of my posts :-)
I was hoping RD Connection Brokers supported NLB as I've also read this is a far superior technology (see below). Here's my architecture thinking:
1. Install the CB role on two individual servers
2. Add NLB to the CBs
3. Follow
here , here and thanks Amy
here for SQL and HA mode (we run Full SQL on an independent server)
Question(s):
a. If SQL is not using AlwaysOn and therefore an SPF, what are the consequences of SQL outage? Does RDS fail completely?
b. If SQL is an SPF, is it worth enabling HA for CB presumably to gain load balancing? This
is a fascinating graph that would suggest little need based on latency for our farm of 20 or so hosts! All servers are on VMware 5.5 therefore inherent redundancy.
We may still choose HA but possibly for MS patching reasons such the farm (minus SQL server) can be patched without downtime...
I found this NLB/RRDNS description useful also
Network Load Balancing is superior to other software solutions such as round robin DNS (RRDNS), which distributes
workload among multiple servers but does not provide a mechanism for server availability. If a server within the host fails, RRDNS, unlike Network Load Balancing, will continue to send it work until a network administrator detects the failure and removes the
server from the DNS address list. This results in service disruption for clients. Network Load Balancing also has advantages over other load balancing solutions—both hardware- and software-based—that introduce single points of failure or performance bottlenecks
by using a centralized dispatcher. Because Network Load Balancing has no proprietary hardware requirements, any industry-standard compatible computer can be used. This provides significant cost savings when compared to proprietary hardware load balancing solutions.
Thanks again
Lea

Round-robin Load balancing doesn't work after one instance reboot

Hi,
          We have two weblogic instances in a cluster. All Http requests are routed through the plug-in for load-balancing(round-robin). One instance in a cluster is shutdown for production maintenance. Now all requests are being forwarded to the second instance. When the first instance is started, only the new users are forwarded to the first instance. The requests from users already logged in are not forwarded to the first instance. Due to this the second instance is over-loaded.
          How to prevent this and make weblogic to distribute the load even after the server restart? This is a production problem and I highly appreciate your responses.
          Thanks,
          Prasad Dantuluri

it worked when I removed the value entered in the "Unicast Broadcast Channel" and restarted the cluster. Don't know what is supposed to go in that field but hey.

WLS Cluster Servlet not "sticky" - continuous round robin (WLS 6.1 SP2)

Similar Messages

Maybe you are looking for