What is the definition of an inactive channel?

Question

On the Lightning Network, after how much time since the last received channel update is a channel considered inactive?

Answer 1

If you look at the channel_update message of BOLT 07 you will find that a node

• MAY create and send a channel_update with the disable bit set to 1, to signal a channel's temporary unavailability (e.g. due to a loss of connectivity) OR permanent unavailability (e.g. prior to an on-chain settlement).
  • MAY sent a subsequent channel_update with the disable bit set to 0 to re-enable the channel.

This means that if you shutdown your node correctly the node or the peer will set the channel to inactive and send a channel message. In particular there is no time(out) encoded to the protocol.

If your node stays online and the TCP-connection stays alive the channel won't become inactive.

If your node dies the peer might send an inactive message. However from the specification it is not clear when and how this will happen.

In general it is considered to be a good practice to shut down your node properly and send out the fact that the channel is inactive so that others won't experience unnecessary routing errors. I think lnd will start collecting statistics about failed routing attempts and locally start to rank nodes by their reliability. While planned downtime is also not as nice as constant uptime it will certainly not downrank your node as much as discovered downtime during routing attempts.

Sorry that this answer might be not as satisfying as you hoped for. Another way to approach this would be to look at the implementations and check when and how they send channel_update messages. For example in c-lightning the send_channel_update function is called 5 times. in Most cases to activate a channel. I can only find 2 cases in which the channel might be broadcasted to be inactive:

static void maybe_send_shutdown(struct peer *peer)
{
    u8 *msg;

    if (!peer->send_shutdown)
        return;

    /* Send a disable channel_update so others don't try to route
     * over us */
    send_channel_update(peer, ROUTING_FLAGS_DISABLED);

    msg = towire_shutdown(NULL, &peer->channel_id, peer->final_scriptpubkey);
    sync_crypto_write(peer->pps, take(msg));
    peer->send_shutdown = false;
    peer->shutdown_sent[LOCAL] = true;
    billboard_update(peer);
}

and:

static void handle_peer_shutdown(struct peer *peer, const u8 *shutdown)
{
    struct channel_id channel_id;
    u8 *scriptpubkey;

    /* Disable the channel. */
    send_channel_update(peer, ROUTING_FLAGS_DISABLED);

    if (!fromwire_shutdown(tmpctx, shutdown, &channel_id, &scriptpubkey))
        peer_failed(peer->pps,
                &peer->channel_id,
                "Bad shutdown %s", tal_hex(peer, shutdown));
    ...

The last one handle_peer_shutdown is the regular shutdown the first one maybe_send_shutdown seems to be called at three points in the code. It can be traced down from there to see if there are any special circumstances at which the channel is broadcasted as inactive.