Services

Default Section

DEFAULT.hard_affinity

This parameter can be used to force a service to run on the same node running other services.

DEFAULT.soft_affinity

This parameter can be used to help a service to run on the same node running other services.

DEFAULT.hard_anti_affinity

This parameter can be used to force a service to avoid execution on the same node running other services.

DEFAULT.soft_anti_affinity

This parameter can be used to help a service try to avoid execution on the same node running other services.

DEFAULT.orchestrate

This parameter is used to specify the daemon behaviour regarding service orchestration

DEFAULT.placement

This setting is used to identify the best node to start the service on.

DEFAULT.constraints

A boolean expression evaluated in order to constrain the service instance placement to nodes where expression is True.

DEFAULT.parents

This parameter is used to store a list of services that must be avail up for the agent daemon to proceed on starting the service.

The parents must be in the same cluster.

DEFAULT.children

This parameter is used to store a list of services that must be avail down for the agent daemon to proceed on stopping the service.

DEFAULT.stonith

A boolean mandating the stale node previously running the service to be shot by the daemon monitor upon failover.

Defaults to false.

DEFAULT.scale

Activates the scaler mode. An integer setting the target number of running instances of this service. To meet this target, the scaler clones itself as slave services. If the scaler topology is a flex, the flex_min_nodes and flex_max_nodes of each slave is automatically adjusted by the daemon.

Defaults to None.

Resource Sections

standby

shared

provision

app.start_timeout

app.stop_timeout

app.check_timeout

app.info_timeout

Node

Node Section

node.maintenance_grace_period

node.rejoin_grace_period

Cluster Section

cluster.secret

Auto-generated.

cluster.nodes

Janitored by the join and leave nodemgr daemon actions.

cluster.name

Defaults to default. Users can set it to a more meaningful value.

Note

Changing that parameter on a joined cluster will break communications until the nodes share the same cluster.name, thus cause a massive split-brain situation. So take care of freezing the cluster nodes before.

cluster.dns

cluster.id

Heartbeat Section

hb.type

hb[type=unicast].addr

hb[type=unicast].intf

hb[type=unicast].port

hb[type=unicast].timeout

hb[type=multicast].addr

hb[type=multicast].intf

hb[type=multicast].port

hb[type=multicast].timeout

hb[type=disk].dev

hb[type=disk].timeout