The baseline topology is a set of nodes meant to hold data. The concept of baseline topology was introduced to give you the ability to control when you want to rebalance the data in the cluster. For example, if you have a cluster of 3 nodes where the data is distributed between the nodes, and you add 2 more nodes, the rebalancing process re-distributes the data between all 5 nodes. The rebalancing process happens when the baseline topology changes, which can either happen automatically or be triggered manually.
The baseline topology only includes server nodes; client nodes are never included because they do not store data.
The purpose of the baseline topology is to:
Avoid unnecessary data transfer when a server node leaves the cluster for a short period of time, for example, due to occasional network failures or scheduled server maintenance.
Give you the ability to control when you want to rebalance the data.
Baseline topology changes automatically when Baseline Topology Autoadjustment is enabled. This is the default behavior for pure in-memory clusters. For persistent clusters, the baseline topology autoadjustment feature must be enabled manually. By default, it is disabled and you have to change the baseline topology manually. You can change the baseline topology using the control script.
Baseline Topology in Pure In-Memory Clusters
In pure in-memory clusters, the default behavior is to adjust the baseline topology to the set of all server nodes automatically when you add or remove server nodes from the cluster. The data is rebalanced automatically, too. You can disable the baseline autoadjustment feature and manage baseline topology manually.
Baseline Topology in Persistent Clusters
If your cluster has at least one data region in which persistence is enabled, the cluster is inactive when you start it for the first time. In the inactive state, all operations are prohibited. The cluster must be activated before you can create caches and upload data. Cluster activation sets the current set of server nodes as the baseline topology. When you restart the cluster, it is activated automatically as soon as all nodes that are registered in the baseline topology join in. However, if some nodes do not join after a restart, you must to activate the cluster manually.
You can activate the cluster using one of the following tools:
Baseline Topology Autoadjustment
Instead of changing the baseline topology manually, you can let the cluster do it automatically. This feature is called Baseline Topology Autoadjustment. When it is enabled, the cluster monitors the state of its server nodes and sets the baseline on the current topology automatically when the cluster topology is stable for a configurable period of time.
Here is what happens when the set of nodes in the cluster changes:
The cluster waits for a configurable amount of time (5 min by default).
If there are no other topology changes during this period, GridGain sets the baseline topology to the current set of nodes.
If the set of nodes changes during this period, the timeout is updated.
Each change in the set of nodes resets the timeout for autoadjustment. When the timeout expires and the current set of nodes is different from the baseline topology (for example, new nodes are present or some old nodes left), GridGain changes the baseline topology to match the current set of nodes. This also triggers data rebalancing.
The autoadjustment timeout allows you to avoid data rebalancing when a node disconnects for a short period due to a temporary network problem or when you want to quickly restart the node. You can set the timeout to a higher value if you expect temporary changes in the set of nodes and don’t want to change the baseline topology.
Baseline topology is autoadjusted only if the cluster is in the active state.
To enable automatic baseline adjustment, you can use the control script or the programmatic API methods shown below:
Ignite ignite = Ignition.start(); ignite.cluster().baselineAutoAdjustEnabled(true); ignite.cluster().baselineAutoAdjustTimeout(30000);
IIgnite ignite = Ignition.Start(); ignite.GetCluster().SetBaselineAutoAdjustEnabledFlag(true); ignite.GetCluster().SetBaselineAutoAdjustTimeout(30000);
To disable automatic baseline adjustment, use the same method with
false passed in: