Removing a Fileserver Role
Describes how to remove the Fileserver role from a node.
About this task
Removing the fileserver role from a node is more complex than removing other roles.
The CLDB tracks data precisely on all fileserver nodes, and therefore you should
direct the cluster CLDB to stop tracking the node before removing the fileserver
role. For a planned decommissioning of a node, use node topologies to migrate data
off the node before removing the fileserver role. For example, you could move the
node out of a live
/data
topology into a
/decommissioned
topology that has no volumes assigned to it, in
order to force data off the node. Otherwise, some data will be under-replicated as
soon as the node is removed. Refer to Node
Topology.NOTE
The following procedure involves halting all data-fabric services on
the node temporarily. If this might disrupt critical services on your cluster,
such as CLDB, migrate those services to a different node first, and then
proceed.The following steps describe how to remove the fileserver role from a node:
Procedure
- Stop the warden, which will halt all data-fabric services on the node. Wait 5 minutes, after which the CLDB will mark the node as critical.
-
Remove the node from the cluster, to direct the CLDB to stop tracking this
node.
You can do this in the Control System GUI or with the
maprcli node remove
command. -
Remove the fileserver role by deleting the file
/opt/mapr/roles/fileserver
on the node. -
Run
configure.sh
on the node to reconfigure the node without the fileserver role. -
Issue the following command to restart Warden on the node:
% service mapr-warden restart
-
Remove any volumes that were stored locally on the node.
You can do this in the Control System or with the
maprcli volume remove
command.