Modifies auto-reprovision settings
couchbase-cli setting-autoreprovision [--cluster <url>] [--username <user>]
[--password <password>] [--enabled <1|0>] [--max-nodes <num>]
Auto-reprovisioning is used in order to prevent data loss in ephemeral buckets during failure scenarios when a node crashes and restarts quickly. Under this scenario auto-reprovisioning ensures that an ephemeral bucket’s replica vBuckets are promoted to active state. This mechanism is not needed for Couchbase buckets because a Couchbase buckets data is persisted to disk and can be loaded back into memory after a node restarts. Ephemeral buckets on the other hand completely lose their data when a node crashes and replicas must be relied upon in order to prevent data loss. If ephemeral buckets are not in use this setting has no effect on the cluster.
Auto-reprovisioning can either be enabled or disabled. If you have ephemeral buckets it is always recommended that auto-reprovisioning is enabled otherwise the cluster will experience data loss if a node crashes or is restarted. Users can also specify the number of nodes that can be auto-reprovisioned before the cluster is rebalanced. An auto-reprovision event occurs if a cluster has at least one ephemeral bucket and a node crashes and restarts. If the number of crashes and restarts exceed the maximum number of nodes that can be auto-reprovisioned then the next crash and restart will result in data loss for all ephemeral buckets in the cluster.
When setting the max nodes parameter, note that specifying the max nodes to be too high could result in cascading node failures. This can happen because when a node fails and restarts the server load for the crashed node is distributed to the rest of the servers in the cluster. This extra load could cause the one or more of the remaining servers in the cluster to become overloaded and unresponsive leading to more failures. On the other hand specifying max nodes to be too small could lead to data loss if there are many failures in the cluster at the same time. How this variable is set depends on the cluster size, workload, and configuration. However, it is always recommended that auto-reprovisioning is at least enabled and that max nodes is set to at least 1.
- --enabled <num>
-
Specifies whether or not auto-reprovisioning is enabled. Set this option to "1" to enable autofailover or "0" to disable autofailover.
- --max-nodes <num>
-
Specifies the maximum number of servers that can be auto-reprovisioned before a rebalance must take place (which resets the count). This parameter must always be set to a number greater than or equal to 1.
To enable auto-reprovisioning and allow up to three nodes be auto-reprovisioned before a rebalance takes place run the following command.
$ couchbase-cli setting-autoreprovision -c 192.168.1.5 -u Administrator \ -p password --enabled 1 --max-nodes 3
To disable auto-reprovisioning run the following command.
$ couchbase-cli setting-autoreprovision -c 192.168.1.5 -u Administrator \ -p password --enabled 0
man:couchbase-cli-setting-autofailover[1]