RE: LMC Clustering Travis Cornwell 19.Jan.09 05:57 AM a Web browser Deployment 6.1.2All platforms
LMC itself provides load balancing capabilities with its clustering. It does not provide "true" high availability clustering. In the event a subordinate node goes down, the cluster will no longer route traffic to that secondary node. In the event the primary node goes down, the cluster can promote one of the subordinate node to become the primary node.
However, this is still not "true" high-availability software. For high-availability, something like HACMP (for AIX), or, Heartbeat (for Linux) would be required. Windows 2003 I know has a range of products / solutions as well. "True" high-availability would be able to monitor the system independent of the software. Meaning, let say an Ethernet cable goes bad on your LMC system, well, LMC itself will not "failover" ... it will attempt to gracefully recover and perhaps give some informative log messages, but it will not "failover". High-availability software would in this instance.
For the load-balancing capabilities, the primary node acts as an entry point, and passes all traffic to subordinate nodes for processing. For the return path of the network traffic, it can be accomplished in one of two ways. All traffic can be rerouted back through the "primary", or, you can select to have "response determined by MNC". For the primary:
LMC Client --> Primary LMC Server --> Subordinate LMC Server --> Primary LMC Server --> LMC Client
Note that the subordinate node does all of the processing / work for that LMC client. If there are multiple subordinate nodes, an algorithm is utilized to distribute the workload. For the "response determined by MNC":
LMC Client --> Primary LMC Server --> Subordinate LMC Server --> LMC Client
This allows direct communicate between the subordinate nodes and the LMC client, reduces the amount of traffic between the primary and subordinate nodes. The entry-point remains the same, but the exit-point differs. If firewall rules will allow this, and, if it can be configured correctly on a test system, I would recommend this form of clustering.
Ultimately, depending on the number of users and system uptime requirements, you may wish to utilize high-availability AND LMC clustering. The high-availability would be utilized on the primary LMC server only, as if the entry-point goes down, won't necessarily matter if the subordinate nodes are still up and running. Given the assumption of a remote database and LDAP server, if the active "primary" node failsover to the standby "primary" node, the failover can read the remote databases and LDAP servers and continue business as usual. End-users will not be disconnected, and likely will only experience minimal "lag" through their VPN Tunnels. The clustering would still be used to distribute traffic workload, and in the event of a failover, if the high-availability is configured properly, would know communication would go back to the standby primary node.
All of this would likely make an excellent Technote. I'll likely write one up in the near future. Please let me know if you have any questions. Thanks.
IBM L2 Support, Lotus Mobile Connect
The postings on this site are my own and do not necessarily represent the positions, strategies, or opinions of IBM.