@Allan_Zimmermann can I install openflow in cluster mode
It depends on what you mean by cluster.
Normally when you talk about clusters, you have active/passive clusters, where only one node is active at the time, or active/active clusters, where both nodes are active and need some way to share storage and state in real time ( normally using quorum disks ). In the first mode, you send all trafic to only one node, in the second you can send traffic to any node. This is relevant for systems that normally would not support running multiple instances, like SQL servers. ( it’s very hard and complex to create an active/active sql cluster … or at least it used to be 
) We use cluster to minimize downtime, not to avoid downtime ( active/passive). The main idear being if one of the servers go down we are quickly up on the other node. We can megiate that using active/active clusters, but depending on they way you implemented that, it might just double the load on the one node, and you might as well just have used an active/passive one.
Another way to scale, would to be use use load balancing. Normally, this means you distribute all client requests across all nodes, either using round robin / ip hash or more intelligent systems like measuring response times. The main difference being with the above you need state across nodes, but with something like web servers, you don’t need to share state. Not always true, like if the web app needs to support sessions, but we can normally solve that by simply using ip hashing.
Now, openflow uses mongodb not a traditional relational database, that actually supports load balencing ( with a twitch ) same goes for rabbitmq, this does support load balancing with the right addons. The api nodes is web apps and they don’t need state, so those can also be load balenced, so that leaves agents ( like nodered ) Most of these do NOT support clustering/load balencing, but if you are using the queue system ( either using amqp nodes or work items ) you can easily scale out but you cannot simply load balence trafic across multiple agents. ( if you started 5 nodereds and set nodered_id to the same id, you can, if the goal is to load balence the http endpoints, but nodered dashboard and the nodered designer does NOT support it, so you will need a agent not part of the load balencing setup to work on the workflows )
So please explain what part of the system you asking about using clustering for, and why ? I don’t see how or why clustering would make any sense, since we can use load balancing or load balancing like options on all elements.
This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.