Normally, silences should be replicated across Alertmanager pods, even after the Alertmanager pods are restarted. However, if persistent storage is not configured and if the Alertmanager pods are restarted at the same time, the silences will not be persisted.

Based on a discussion in Slack, we should add content to the existing Alertmanager docs to explicitly document this behavior and update the persistent storage content to list this as an advantage of configuring persistent storage. In other words, make sure that you configure persistent storage for Alertmanager so that:

• silences persist
• you avoid duplicate notifications on cluster and node upgrades