NAME

drbd-graceful-disconnect.service - Graceful DRBD disconnect on network down, reconnect on network up

SYNOPSIS

drbd-graceful-disconnect.service

DESCRIPTION

This service manages DRBD network connections in response to network lifecycle events.

On stop (network-online.target going down), it disconnects all currently connected DRBD peers. Unlike an unexpected loss of network connectivity, graceful disconnect allows secondary nodes to mark their disks as Outdated during the disconnect two-phase commit. The benefit is that remaining partitions containing a primary node can then keep quorum. The list of disconnected peers is saved to /run so they can be reconnected on start.

On start (network-online.target coming up), it reconnects all peers that were previously disconnected by the stop action.

Peers that are deliberately StandAlone are not touched. Only connections that are currently active in the kernel are affected.

This unit is ordered after network-online.target: when the network goes down, this service will be stopped before the network is torn down.

Resources that are explicitly managed by other systemd services should be stopped there. So this is also ordered before resource-agents-deps.target and drbd.service and the per-resource drbd@.service templates, which makes it be stopped only after those are stopped.

If you have some other foo.service that starts and stops DRBD resources, you should add a drop-in /etc/systemd/system/drbd-graceful-disconnect.service.d/foo.conf

[Unit]
Before=foo.service

(Or the inverse, add an After=drbd-graceful-disconnect.service to your foo.service).

If you for some reason want to opt-out of this service, you can still systemctl mask it.

SEE ALSO

drbd-graceful-down.service(7)

AUTHORS

LINBIT HA Solutions GmbH https://linbit.com