License

A cluster uses one license file that lists every Webserver of the cluster. This page says how to get it, when it must be replaced, and what the license allows when only one Webserver is left.

One license for the cluster

Run the fingerprint tool on each Webserver and send all the tokens; install the file you receive on each of them. The quota (monitoring points, agents) is that of the cluster, not of each node. The witness needs no license.

The first server listed in the license has two extra rights: it may run standalone under the cluster license, and it is never subject to the single-Webserver rule below. Ask for the server that holds your data to be listed first.

When a new license is needed

What changes	New license?
A Webserver is replaced by a new machine	Yes. The license lists the hardware of each Webserver, so the new machine must be in it. Install the new file on both Webservers.
The same machine is reinstalled, or gets a minor hardware change	No.
The witness is replaced or moved	No. It needs no license.

While you wait for the new file, the remaining Webserver keeps running under the current license: the grace period below covers that time.

A cluster reduced to one Webserver

A cluster license covers several servers so that they form one Webserver. If the cluster has had a single Webserver node for longer than the grace period of the license (14 days unless your license says otherwise), everything keeps running and alerting, but no new monitor can be added until a second licensed node is back in the cluster.

The grace period covers a first installation, a failed node being replaced and an emergency recovery.
The rule never applies on the first server listed in the license.
The cluster panel shows the date from which it will apply.

If the first server of the license is the one that failed

Losing the first server is never an outage. The surviving Webserver cannot simply be switched to standalone (that is reserved to the first server), but it runs alone as a cluster of one:

On the surviving server, stop the service and run webserver ha recover confirm, then start the service. It is the active Webserver with all its data, under the same license.
Everything works normally for the grace period (14 days unless your license says otherwise).
Before the grace period ends, either bring a second licensed server into the cluster, or ask for a new license. A replacement for the failed server is new hardware, so it needs a new license in any case; ask for the surviving server to be listed first and it is no longer subject to the rule.

If the grace period runs out before that, monitoring and alerting continue unchanged. Only the creation of new monitors is suspended until a second licensed server is back or the new license is installed.

Running standalone with a cluster license

Removing the HA block makes a node standalone again (see Operations). With a cluster license this is allowed on the first server listed in the license only. On another server the Webserver refuses to start standalone; use webserver ha recover instead.