Taking a Conferencing Node out of service

If you need to take a Conferencing Node out of service, you must first put it into maintenance mode to ensure that it is not hosting any conferences.

When a Conferencing Node goes into maintenance mode, it will not accept any new calls or registrations. Any existing calls on that Conferencing Node will not be affected (this applies to all call processing i.e. handling signaling, media proxying and transcoding). After all existing calls have terminated, the Conferencing Node will still be live and contactable but will not be handling any calls.

A Conferencing Node will go into maintenance mode if:

  • The system administrator elects to put the Conferencing Node into maintenance mode via the Pexip Infinity Administrator interface (see Manually placing a Conferencing Node into maintenance mode below).

    This setting will persist after a reboot.

  • The Pexip Infinity platform is being upgraded, during which time each Conferencing Node in turn is automatically put into maintenance mode and upgraded. For more information, see Upgrading the Pexip Infinity platform.
  • The Conferencing Node has been installed on a system that does not meet the CPU instruction set requirements. In such a case:

    • after initial deployment, the following message will appear in the admin log: CPU instruction set is not supported; system will be placed in maintenance mode
    • each time the Conferencing Node rejects a call, the following message will appear in the admin log: Message="Participant failed to join conference." Reason="System in maintenance mode"
    • any manual changes to the Enable maintenance mode setting will have no impact - the Conferencing Node will remain in maintenance mode regardless of this setting.

    For more information on how to resolve this, see Troubleshooting the Pexip Infinity platform.

  • The Conferencing Node is running on a VM in Azure which has a scheduled maintenance event.

Note that when a Conferencing Node is in maintenance mode, it reports a media load of 100%. This is to indicate that there is no current capacity available.

Manually placing a Conferencing Node into maintenance mode

When putting multiple nodes into maintenance mode, to ensure maximum availability for registration data and other data that may have replicas stored in the distributed database on that node, and to ensure maximum service availability, we recommend that you:

  • Activate maintenance mode on one node at a time.
  • Wait a couple of minutes before putting the next node into maintenance mode.

This allows registrations and any relevant distributed data to gradually be migrated to other nodes.

To manually put a Conferencing Node into maintenance mode:

  1. Go to Platform > Conferencing Nodes.
  2. Select the Conferencing Node(s).
  3. From the Action menu at the top left of the screen, select Enable maintenance mode and then select Go.

    While maintenance mode is enabled, this Conferencing Node will not accept any new conference instances.

  4. Wait until any existing conferences on that Conferencing Node have finished. To check, go to Status > Live View.

When all conferences on the Conferencing Node have finished you can take it out of service.

Long term maintenance

If you intend to take one or more nodes offline for a long period of time (days or weeks) then you should consider temporarily deleting those nodes altogether — and redeploying them afresh when they are next needed. This will avoid distracting connectivity alarms being raised for those nodes that might make it harder for the administrator to spot other "unexpected" alarms in the meantime.

Detecting maintenance mode

Load balancers can use the https://<node_address>/api/client/v2/status REST API command to check whether a Conferencing Node is in maintenance mode.