Document mattermost_cluster_reliable_fallback_tcp

Author: agarciamontoro

source/administration-guide/scale/performance-monitoring-metrics.rst

@@ -37,6 +37,7 @@ Cluster metrics
 - ``mattermost_cluster_cluster_health_score``: A score that gives an idea of how well it is meeting the soft-real time requirements of the gossip protocol.
 - ``mattermost_cluster_cluster_requests_total``: The total number of inter-node requests.
 - ``mattermost_cluster_cluster_event_type_totals``: The total number of cluster requests sent for any type.
+- ``mattermost_cluster_reliable_fallback_tcp``: The total length in bytes of the SendBestEffort calls (UDP) that had to fallback to SendReliable (TCP) because of the message length. This is a histogram with 8 buckets, from 32KiB to 4MiB. Ideally, the count of this metric should be always 0, but plugins may be misusing the websocket API and attempting to send large messages via UDP; in that case, plugins should be fixed to set `ReliableClusterSend <https://github.com/mattermost/mattermost/blob/4fe11a7d5d14616f3c31363762617ddfd99895df/server/public/model/websocket_message.go#L155-L157>`__ to true.
 
 Database metrics
 ~~~~~~~~~~~~~~~~