Mnesia

Description

The Mnesia adapter stores LiveView state on the server using Mnesia through the memento wrapper. Add it to your dependencies to use this adapter. The table is stored in memory. It uses Mnesia replication to make state recovery easier, and rolling redeploys possible. State is stashed into native Mnesia table copies, recovered from the local copy on reconnect, and cleared when you reset the stash.

Replication

The adapter always configures native Mnesia table copies on connected nodes and relies on Mnesia replication for writes/deletes.

Recovery is local because every serving node is expected to have the table copy already.

When to use

Choose the Mnesia adapter when:

• BEAM native solution: You want a fully server-side solution that leverages the BEAM's distributed capabilities without relying on external services like Redis.
• State durability (Rolling Redeploys): You need stashed state to survive rolling server restarts and redeployments. Thanks to Mnesia's replication, the state persists as long as at least one node in the cluster remains active. (Note: A simultaneous shutdown of the entire cluster will clear the state).
• Clustering: You have a clustered application where client can connect to different nodes.

Resetting the stash

Stashed state is automatically cleared after the TTL passes, provided the process owning the state is dead. If the process is still alive, the delete_at time gets bumped by the TTL.

State can also be cleared manually by calling LiveStash.reset_stash/1.

Configuration

Activating the adapter

Remember to define adapters you would like to activate in your config.exs file.

config :live_stash, adapters: [LiveStash.Adapters.Mnesia]

Expiration (TTL)

Stashed data in server mode has a Time-To-Live (TTL) to prevent stale state from persisting indefinitely. You can adjust this using the :ttl option.

Default: 300 seconds (5 minutes)

use LiveStash, adapter: LiveStash.Adapters.Mnesia, ttl: 60, stored_keys: [:count]

Cleanup interval

Determines how often the background task runs to remove expired state records from the Mnesia table.

Default: 60_000 ms (1 minute)

To override this, add the following to your config/config.exs:

config :live_stash, adapters: [LiveStash.Adapters.Mnesia], mnesia_cleanup_interval: 60_000

Split brain resolution strategy

By default, in reaction to :inconsistent_database event from Mnesia, the lexicographically larger node yields and copies state from the smaller node (nodeA > nodeB). You can pass mnesia_auto_heal: false and implement your own strategy.

config :live_stash, adapters: [LiveStash.Adapters.Mnesia], mnesia_auto_heal: false

Security

By default, LiveStash uses a hardcoded default secret ("live_stash") to secure your data. For production environments, it is highly recommended to tie the stash to a specific user session to prevent tampering or data leakage.

You can do this by providing a :session_key. LiveStash will extract the value from the connection session securely hash it (SHA-256) to use as the operational secret. If you provide the key and it is not present in the session, Argument Error will be raised.

In Mnesia mode, this operational secret is used as part of the record ID for your stashed state.

use LiveStash, adapter: LiveStash.Adapters.Mnesia, session_key: "user_token", stored_keys: [:count]