Behavior of nodegroups
Due to the interoperability of nodegroups with different NetScaler features and entities, there are some behavioral aspects to be noted. Nodes in a nodegroup can also be backed up. Read on for more information.
General behavior of a cluster nodegroup
- A nodegroup that has entities bound to it cannot be removed.
- A cluster node that belongs to a nodegroup with entities bound to it, cannot be removed.
- A cluster instance that has nodegroups with entities bound to it, cannot be removed.
You cannot add an entity that has a dependency on another entity that is not part of the nodegroup. If you need to do so, first remove the dependency. Then, add both the entities to the nodegroup and reassociate the entities.
- Assume you have a virtual server, VS1, whose backup is virtual server VS2. To add VS1 to a nodegroup, first make sure that VS2 is removed as the backup server of VS1. Then, bind each server individually to the nodegroup, and then configure VS2 as the backup for VS1.
- Assume you have a content switching virtual server, CSVS1, whose target load balancing virtual server is LBVS1. To add CSVS1 to a nodegroup, first remove LBVS1 as the target. Then, bind each server individually to the nodegroup, and then configure LBVS1 as the target.
- Assume you have a load balancing virtual server, LBVS1, that has a policy which invokes another load balancing virtual server, LBVS2. To add either one of the virtual servers, first remove the association. Then, bind each server individually to the nodegroup, and then reassociate the virtual servers.
- You cannot bind an entity to a nodegroup that has no nodes and that has the strict option enabled. Consequently, you cannot unbind the last node of a nodegroup that has entities bound to it and that has the strict option enabled.
- The strict option cannot be modified for a nodegroup that has no nodes but has entities bound to it.
Backing up nodes in a nodegroup
By default, a nodegroup is designed to provide back up nodes for members of a nodegroup. If a nodegroup member goes down, a cluster node that is not a member of the nodegroup dynamically replaces the failed node. This node is called the replacement node.
Note: For a single-member nodegroup, a backup node is automatically preselected when an entity is bound to the nodegroup.
When the original member of the nodegroup comes up, the replacement node, by default, is replaced by the original member node.
From NetScaler 10.5 Build 50.10 onwards, however, the NetScaler appliance allows you to change this replacement behavior. When you enable the sticky option, the replacement node is retained even after the original member node comes up. The original node takes over only when the replacement node goes down.
You can also disable the backup functionality. To do this, you must enable the strict option. In this scenario, when a nodegroup member goes down, no other cluster node is picked up as a backup node. The original node continues being part of the nodegroup when it comes up. This option ensures that entities bound to a nodegroup are active only on nodegroup members.
The strict and sticky option can be set only when creating a nodegroup.