Hi,
I have 2 servers Server A ( Deployment Manager, IIS Server, App Server) and 
Server B ( IIS and App Server). Both app server are federated into Deploymen
t manager profile. I have configured IIS to work with WAS. It works perfectl
y on Server A and I have co
nfigured IIS and WAS same way on Server B . Through deployment manager web s
erver plugins are successfully generated I have verified by opening the plug
in file on each server, it contains URI for App being installed, I have chec
ked on both servers). The I
ssue I am having is when I shut down WAS on Server A and try to hit the serv
er with its host e.g htt://serverb/snoop , When I look at the plugin log on 
server B, it fails with the error " ws_common: websphereGetStream: Failed to
 connect to app server on h
ost serverA" . When I look at the plugin log on Server B, it showing me that
 its trying to connect to WAS on Server A instead of WAS on Server B, which 
is down so it fails. If someone know what's going on here, it would be highl
y appreciated.

Thanks,
DV
dvpunia@decisionlabs.com

[ Post a follow-up to this message ]

DVPunia@decisionlabs.com wrote:
> Hi, I have 2 servers Server A ( Deployment Manager, IIS Server, App
> Server) and Server B ( IIS and App Server). Both app server are
> federated into Deployment manager profile. I have configured IIS to
> work with WAS. It works perfectly on Server A and I have configured
> IIS and WAS same way on Server B . Through deployment manager web
> server plugins are successfully generated I have verified by opening
> the plugin file on each server, it contains URI for App being
> installed, I have checked on both servers). The Issue I am having is
> when I shut down WAS on Server A and try to hit the server with its
> host e.g htt://serverb/snoop , When I look at the plugin log on
> server B, it fails with the error " ws_common: websphereGetStream:
> Failed to connect to app server on host serverA" . When I look at the
> plugin log on Server B, it showing me that its trying to connect to
> WAS on Server A instead of WAS on Server B, which is down so it
> fails. If someone know what's going on here, it would be highly
> appreciated.


This is normal behavior. A webserver plugin will load balance across the
cluster. Does it fail one time only and then only access the "up"
server, or does it fail repeatedly? It should notice that the appserver
is down and stop sending work to it.

[ Post a follow-up to this message ]

Hi Paul,
Thanks for your response. There is no clustering on these servers yet. These
 are 2 individual servers on 2 different machines which are federated throug
h Deployment Manager. Does it fail repeatedly? yes it does, it seems to alwa
ys go to server 1 no matter
what.

Regards,
DV
dvpunia@decisionlabs.com

[ Post a follow-up to this message ]

DV wrote:
> Hi Paul, Thanks for your response. There is no clustering on these
> servers yet. These are 2 individual servers on 2 different machines
> which are federated through Deployment Manager. Does it fail
> repeatedly? yes it does, it seems to always go to server 1 no matter
> what.

Well, if there's no clustering, it will go to whichever server is
specified in the plugin config file.

[ Post a follow-up to this message ]

When the application is installed using deployment manager both servers are 
selected at deployment. So after saving the configuration using was admin co
nsole. Application is installed on both servers. So When I look at the plugi
n, it shows both servers in
there but when I hit the server 2 with the explicit server name (e.g http://server
2/snoop), it connects to the web server on server 2 and then the web server 
on server 2 tries to connect to was on server 1 instead of the server 2. Thi
s info is logged int
o plugin log file. So how I can avoid that. Realisticly if I hit server2 wit
h its host name, it should go to server 2.

Question: When 2 servers are federated using deployment manager, are they au
tomatically treated like clusters because by looking in the plugin it shows 
"round robin".


Attached is the plugin file.
Thanks.
DV
dvpunia@decisionlabs.com

[ Post a follow-up to this message ]

DV wrote:
> When the application is installed using deployment manager both
> servers are selected at deployment. So after saving the configuration
> using was admin console. Application is installed on both servers. So
> When I look at the plugin, it shows both servers in there but when I
> hit the server 2 with the explicit server name (e.g
> http://server2/snoop), it connects to the web server on server 2 and
> then the web server on server 2 tries to connect to was on server 1
> instead of the server 2. This info is logged into plugin log file. So
> how I can avoid that. Realisticly if I hit server2 with its host
> name, it should go to server 2.

No, that's not how the plugin works. If you want the web server on
server 1 to only go to the appserver on server 1, you need to only have
appserver 1 in the plugin config on that box. You're going to have to
create custom plugin files.
>
> Question: When 2 servers are federated using deployment manager, are
> they automatically treated like clusters because by looking in the
> plugin it shows "round robin".

No, if they were treated like clusters it would load balance.

[ Post a follow-up to this message ]

Hi Paul,
How do you create custom plugin files, I know you can create custom profile 
but never knew you could create custom plugin files, do you mean manually, I
 hope not. One thing I am not clear is, When I installed WAS on Server1, I c
reated 2 profiles, one for
Deployment Manager and other for AppSrv01. After creating 2 profiles, I inst
alled the Http plugin using install wizard and did the test to see if plugin
 was installed/configured. It was all good because I was able to hit the sno
op servlet wihtout specifin
g the port 9080. That shows me it was good. So my question is when you insta
ll the plugin, how do you know what profile its associating it with, I am as
suming it would be , AppSrv01, correct.

Thanks,
DV
dvpunia@decisionlabs.com

[ Post a follow-up to this message ]

DV wrote:
> Hi Paul, How do you create custom plugin files, I know you can create
> custom profile but never knew you could create custom plugin files,
> do you mean manually, I hope not.

Yes, I mean manually. but you'd be better off just defining a cluster
and deploying the application to the cluster. Then it would nicely load
balance across the machines. BTW, normally people do not run the
webserver on the same machine as the appserver, at least not in
production and performance test environments.

See this link for everything you want to know about the plugin config file:

http://publib.boulder.ibm.com/infoc..._plugincfg.html

and also see:

http://publib.boulder.ibm.com/infoc...wsv_plugin.html

[ Post a follow-up to this message ]

Hi Paul,
I truly appreciate your responses. I am not using cluster there is a reason 
for not using it for now may be in future that might be the direction to tak
e and the reason I am using webserver and WAS server on the same machine bec
ause they are strong machin
e and hardware is an issue, don't have too many machines.  I already gone th
rough the links you have mailed me. I think the issue is when using deployme
nt manager to deploy the application and the app is installed on both server
s and then you generate the
plugin , it automatically adds both server entries in the plugin in tags => 
[ <serverCluster Name="server1"> & <Route ServerCluster="server1"> and <
serverCluster Name="server2"> & <Route ServerCluster="server2" ] , eventhoug
h  they are not cluster but it
puts both server entires anyway. There must be a better way to do this, for 
now I have to manually remove the entries of server1 from server2 plugin fil
e and vice versa.
If you know a better way to do this , will truly appreciate your response, m
ay be through script.

Once again thanks for sharing your thoughts.

DV
dvpunia@decisionlabs.com

[ Post a follow-up to this message ]

Hi,

I think your problem is session stickiness, we can achive this with implemen
ting cluster. There is a work around to over come this issue.

Application servers > <server1> > Web container > Custom Properties add the 
below entries,
Name                cloneid      description
----                 -------     -----------
HttpSessionCloneId  11mqxyzvk     server 1

The same clone id has to configure in the plugin-cfg.xml file.

You need to add this cloneid for both server instances, then the session sti
ckiness problem will be solved.

Follow this if it is helpful.

Thanks
Jai Nadi

[ Post a follow-up to this message ]