Kregedit entries for cluster

Question

Can someone show me what a properly configured cluster looks like in
kregedit? I have just installed two instances of iAS on the same machine -
just to prototype a cluster. My kregedit entries for each are identical:
SOFTWARE/iPlanet
Application Server
Clusters
NASCluster
MaxBackups=1
MaxHops=1
MaxSyncHeartBeat=10
SyncHeartBeatInterval=30
SyncTimerInterval=60
SyncServers
169.170.40.2:60818=0
169.170.40.2:61818=1
I started 169.170.40.2:60818 first and then 169.170.40.2:61818. They both
started up fine and requests to the single web server for both these app
servers correctly sent requests to 169.170.40.2:60818, the primary. When I
stopped the primary I got the following in the kjs logs for
169.170.40.2:61818:
[21/Aug/2001 16:06:07:4] warning: DSYNC-045: We(0xa9aa2802:61820), lost
connection to the Primary, 0xa9aa2802:60818
[21/Aug/2001 16:06:07:4] error: CONN-003: socket receive error (RecvBuffer
1) (conn 0x78a590, sock 33, host 0xa9aa2802, port 0)
[21/Aug/2001 16:06:07:4] info: PROT-001: RecvBuffer failed during
GXRunConn::Main()
[21/Aug/2001 16:06:17:4] info: PROT-006: new connection established
[21/Aug/2001 16:06:21:5] info: DSYNC-038: We(0xa9aa2802:61820), are told
that new Primary is 0xa9aa2802:61818
After all this, it appears that the web server isn't hitting
169.170.40.2:61818 and subsequent requests for the fortune servlet give the
following:
ERROR: Could not connect to the requested service. Check if it has been
started.
Thanks,
Lincoln

Answer 1

Hi Lincoln,
Your cluster is absolutely fine there is nothing wrong with your
settings, and as the kjs logs says that the backup server is also fine
(taking over as primary). Actually the problem is with the application
which you are trying to access in a distributed manner.
By default the fortune application is not registered as distributed,
but if you want to have it distributed, than please make these changes
in the iAS registry:
- run kregedit
- go to iplanet/Application Server/6.0/J2EE-Module/fortune node
than
- insert your domain name value in session-domain key.
- modify the session-impl key value to distributed from lite.
- insert the session-path key value as / (slash)
- ensure that the key session-dsync-type is dsync-distributed
at the end, your key values should look like this -
- session-domain=your.domain.name
- session-dsync-type=dsync-distributed
- session-impl=distributed
- session-path=/
Note: you should make these changes in the registry of all the iAS
servers that are participating in your cluster, or you can redeloy the
fortune app. on all the servers, with appripriate deployment
descriptor.
Change the load balancing method for all iAS servers to Round
Robin(Web connector Driven). You can do it from iAS admin tool, go to
Load Balancing tab and change it.
Than restart the two web servers(iws) and the two iAS servers in the
cluster.
Note: start the highest priority server first.
Now you should be able to have the distributed session with fortune
also.
The clustering and the distributed session failover can be better seen
through the clustering examples/samples provided with iAS, like
clustering-servlet and clustering-jsp.
Please see the page:
http://<hostname.domainname>:<http-port>/ias-samples
for all the sample applications tha comes with iAS and
http://<hostname.domainname>:<http-port>/ias-samples/cluster/docs/index.html
for cluster related sample application.
If you have any further doubts, please mail me back.
Regards,
Sanjeev.
Developer Support Group iAS
Lincoln wrote:
>
Can someone show me what a properly configured cluster looks like in
kregedit? I have just installed two instances of iAS on the same machine -
just to prototype a cluster. My kregedit entries for each are identical:
SOFTWARE/iPlanet
Application Server
Clusters
NASCluster
MaxBackups=1
MaxHops=1
MaxSyncHeartBeat=10
SyncHeartBeatInterval=30
SyncTimerInterval=60
SyncServers
169.170.40.2:60818=0
169.170.40.2:61818=1
I started 169.170.40.2:60818 first and then 169.170.40.2:61818. They both
started up fine and requests to the single web server for both these app
servers correctly sent requests to 169.170.40.2:60818, the primary. When I
stopped the primary I got the following in the kjs logs for
169.170.40.2:61818:
[21/Aug/2001 16:06:07:4] warning: DSYNC-045: We(0xa9aa2802:61820), lost
connection to the Primary, 0xa9aa2802:60818
[21/Aug/2001 16:06:07:4] error: CONN-003: socket receive error (RecvBuffer
1) (conn 0x78a590, sock 33, host 0xa9aa2802, port 0)
[21/Aug/2001 16:06:07:4] info: PROT-001: RecvBuffer failed during
GXRunConn::Main()
[21/Aug/2001 16:06:17:4] info: PROT-006: new connection established
[21/Aug/2001 16:06:21:5] info: DSYNC-038: We(0xa9aa2802:61820), are told
that new Primary is 0xa9aa2802:61818
After all this, it appears that the web server isn't hitting
169.170.40.2:61818 and subsequent requests for the fortune servlet give the
following:
ERROR: Could not connect to the requested service. Check if it has been
started.
Thanks,
Lincoln