This document describes useful information when administering a bridge. If you are looking for information on how to set up the bridge, please see Bridge Setup.
Typically the information given in Bridge Setup is good enough for small to medium sized bridges, but if you expect to handle extremely heavy IRC or Matrix traffic then it might be looking at tweaking some of these options.
It should be noted that often the homeserver you have connected to the bridge will play a greater role in the percieved performance of your bridge, as it is usually the bottleneck in either handing (federated) traffic towards the bridge, or persisting and federating traffic from the bridge to users.
It is strongly advised that if you are suffering from performance issues, you should identify if there are problems with your homeserver first.
This setting handles the "debouncing" of quits when the server sees an extreme amount of QUIT events from the IRC server. IRC servers often suffer from netsplits which manifest as many QUITs. The IRC bridge will handle one QUIT per room, so 5 users quitting from 5 rooms would manifest as 25 events. This can quickly overwhelm the bridge.
The quit debouner is often overkill for smaller bridges, but if you find that the bridge becomes overwhelmed and unresponsive after a netsplit then it enabled.
Typically it's wise to leave this setting on by default, as populating the memberlists on both sides of the bridge leads to a more pleasant experience for users. However as the setting requires the constant adjustment of the member lists on both sides of the bridge it can be more intensive on homeserver resources. You can also adjust the membership settings of individual rooms or channels to lessen the effect.
The bridge supports hot-reloading of the configuration file by sending a
SIGHUP signal. Some configuration
keys will not be reloaded as they are required to be static to avoid bridge instability. Unsupported keys are
marked in config.sample.yaml.
Hot reloading is useful as restarting the bridge will drop all IRC connections, so it's worth using this
method to avoid disruption.
Typically the process is as follows.
$ ps -A | grep 'node' 31960 pts/7 00:00:13 node # and then send the SIGHUP signal $ kill -SIGHUP 31960
The logs will then mention
Bridge config was reloaded, applying changes which confirms
that the reload has taken place.
Enforcing Matrix users to be connected to IRC
When configured to do so, the IRC bridge typically tries to join all Matrix users to the IRC channels to avoid Matrix users being able to read a conversation without being visible to IRC users. However since it is not always possible to ensure this happens in a timely manner, there is a safety net feature.
Administatators can choose the default behaviour of allowing messages to continue to be bridged to the room (potentially leaking history) or enforcing strict rules to ensure that all Matrix users are joined before anyone can read messages. This can be enabled by setting
... membershipLists: global: ircToMatrix: requireMatrixJoined: true
in the config. Users can choose to disable this on a per-room basis by modfiying their room config options, if the bridge permits it.
Metrics / Grafana
The bridge includes a prometheus compatible metrics endpoint which can be used to inspect the state of the bridge. The repository also includes a grafana dashboard; more information can be found in GRAFANA.md.
The Debug API
The Debug API allows you to perform administrative actions on the bridge such as killing a IRC connection, inspecting connected users or unbridging a room. You can learn more by reading the Debug API documentation.