Cluster nodes showing offline status on SMP 3.0 SP5 for Linux machine

[Former Member]

We have installed SMP 3.0 SP5 on 2 linux machines(rhel 6.4)  in a cluster environment with MSSQL 2012 database.

We are able to use these servers and applcation is also syncing successfully.

The two servers can ping each other as well

When we login to the SMP admin portal, we notice that on the Cluster page, only one Vluster node is ONLINE and the other OFFLINE and vice-versa.

For this we have tried, Troubleshooting guides mentioned in the guide, but have not succeeded.

such as,

deleting

In<SMP_HOME>\Server\configuration\com.sap.mobile.platform.server.launcher\fixed-sys.properties, remove the following line:  

java.net.preferIPv4Stack=true

Adding the binding ip in props.ini file of both server.

In the dtabase we can see entries for this in SMP_CLUSTER_MEMBER table.(attached)

We need help  in setting this correctly.

Also, sometimes the cluster page is blank as seen in screenshot.

Attaching the screenshots.

[agentry_src]

Discussion successfully moved from SAP for Mobile to SMP Developer Center as the more appropriate venue for this topic.

Regards, Mike (Moderator)

SAP Technology RIG

[daniel_derron]

Hi there

SAP got back to me with the solution below. Our first test's on DEV where successfull.

Best regards

Daniel

SAP Active Global Support Engineering feedback : We have another incident for the same issue. The primary issue of Admin UI not showing all nodes is resolved. There is still a pending issue that the UI shows the wrong port number, which we are looking into.

The workaround was to modify the file Server/configuration/com.sap.mobile.platform.server.cluster/SMPjGroups.xml and replace it contents with that of the the attached SMPjGroups.new.xml.

Please modify the file strictly as follows :

1. Ensure all cluster nodes are running

2. Open the file in an editor and modify its contents in place.

3. Save it.

4. The changes will be propagated to the other nodes.    Monitor the other nodes to wait until the changes are reflected on all nodes.

5. Restart the server on all nodes.

Important :

1. Do NOT modify the file on a node while the server is stopped. It will be reverted ( or worse, deleted ) when the server starts.

2. Modify the file in an editor, do NOT copy the file over. This may give it an older timestamp and result in strange behavior.

With the above configuration, the following firewall configuration is required. More number of ports will be required to be open if there are are more nodes in the cluster.

This configuration has been tested with 2 nodes.

iptables -A INPUT -p tcp --dport 7900:7960 -j ACCEPT iptables -A OUTPUT -p tcp --sport 7900:7960 -j ACCEPT iptables -A INPUT -p tcp --dport 20:22 -j ACCEPT iptables -A OUTPUT -p tcp --sport 20:22 -j ACCEPT iptables -A INPUT -p tcp --dport 8080:8083 -j ACCEPT iptables -A OUTPUT -p tcp --sport 8080:8083 -j ACCEPT iptables -A INPUT -p tcp --dport 1717 -j ACCEPT iptables -A OUTPUT -p tcp --sport 1717 -j ACCEPT iptables -A INPUT -p tcp --dport 1024:2500 -j ACCEPT iptables -A OUTPUT -p tcp --sport 1024:2500 -j ACCEPT service iptables save iptables -L -n

Engineering is working with the team to provide complete documentation.

[j_raymakers]

Hi Daniel,

Many thanks for your quick reply.

Allthough I have some more questions.

Those port ranges mentioned above...

7900:7960

20:22

8080:8083

1717

1024:2500

do they need be opened on the firewall between the 2 application server nodes

OR

Between Server nodes and Database server

OR

Both

Thanks in advance!

Jacco Raymakers

[daniel_derron]

Hi Jacco ...at least between both application Servers! We are currently working on this with SAP and as soon we have the correct answer I will post it here. Best regards Daniel

[former_member220966]

We have an OSS note open with SAP for the same issue. As a workaround, we opened a large range of ports b/w the 2 nodes, so the bind port(which is generated randomly) always falls within this range.

Now we see both nodes as online - this is not the solution though, only a work around. Try giving this a shot :-).

Cheers,
Abhinav

[j_raymakers]

Hi Abhinav,

We are facing the same issue.

We have 2 application nodes and 1 separate DB host.

The 2 application nodes are windows 2008 environments. The database server is a Linux environment.

The 2 nodes are synchronizing data but status is online / offlline, as descibed above.

Question: did you get feedback from SAP? And what was the answer?

Thanks in advance.

Jacco Raymakers.

[j_raymakers]

PS: SMP 3.0 SP07 installed.

[daniel_derron]

Hi,

from our Point of view it has to do with kind of a blocked intercommunication between the mobile Servers. when you turn off the Firewall, both Servers show the correct Status (online). So there must be a communication between both Servers. You can check this also when you start MS1 first and then MS2. The 2nd Server has always longer until he is ready. Vice versa it is the same. We also have all the necessary ports which are in our case 1521, 8080,8081,8082 and 8083 in the firewallsettings. But it still works only when the Firewall is completely switched off.

Regards, Daniel

[Former Member]

Hi,

We tried the cluster installation for Windows as well, and we found the same error.

What exactly should be the value of bind address?

Is it a common ip ? Should this ip and port be same for both the Server Nodes?

In need your suggestions!

Regards,

Suraj

[Former Member]

Hi people,

Anyone's got an idea about why this is happening?

Regards

Suraj

[kirankola]

Hi Suraj,

I see your DB entry as local host, is that correct?

Regards,

Kiran

[Former Member]

Hi Kiran,

That too is actually incorrect.

It takes details of 2nd server as localhost which is incorrect.

The first server detail is correct.

Regards,

Suraj