Configuring Cluster Resources and Properties
A resource is anything managed by the cluster. Resources are represented by resource scripts.
There are four main types:
OCF – open cluster framework
systemd – to unit files of systemd – you can take these out of systemd and will be run by the cluster instead
heartbeat – this was the old communication system for clustering, avoid if you can. Most now replaced by ocf scripts
stonith – these are scripts for stonith devices
The resource config lives in the CIB cluster information base
Three main types of resources:
- primitive – a single resource that can be managed, usually only needs to start once eg for an ip address
- clone – should run on multiple nodes at same time
- multi state – (master/slave), this is a special form of clone. applies to specific resource only, usually ones involving master-slave
group resource type – makes it easier to manage resources by grouping related resources together, ensures they can be started /stopped together and can be related and linked, for starting/stopping sequence
Resources that are part of the same resource group:
- Start in the defined sequence.
- Stop in the reverse order.
- Always run on the same cluster node.
They can be a group of primitives –
this is when a resource will go down after original situation is restored. this defines what should happen to resource when a node has been restored to the cluster after fencing.
eg resource should migrate back to the original node, or to stay where it is.
but- generally its best to avoid resources migrating from node to node.
find / -name IPaddr2
here we have
.isolation – for docker wrappers
under heartbeat we have a whole long list of scripts eg IPaddr2
note IPaddr2 is for the ip suite of network commands the IPaddr is for ifconfig. you should only be using IPaddr2 nowadays.
under crm shell
this shows you the same as above
also there is the info command
this displays the shell script meta data for IPaddr2 shell script
crm configure primitive newip ocf:heartbeat:IPaddr2 params ip_ip address op monitor interval-10s
crm resource show newip
this is specific to the crm command
crm_mon will show you the list of your current active resources live and running on your cluster
allows you to query the cib
CAUTION: these are dangerous, use with care!
Resources have to be related to each other, this can be done by creating resource constraints:
Location: on which node/s the resource should run – can be done positively or negatively with scores
Colocation: with which resource a resource should run
Order: after/before which resourse
[root@centos1 corosync]# pcs constraint show
Typically a score is used, between
INFINITY: must happen, and
-INFINITY: may not happen
Intermediate values: expresses greater or lesser wish to have it happen or not
To ensure a certain action is never performed, use a negative score. Any score smaller than 0 will ban the resource from a node.
crm migrate / pcs resource move – these also enforce INFINITY resource constraints, you will need to remove this this using
crm resource unmigrate /pcs resource clear
NOTE: -INFINITY on a location constraint will NEVER run the resource on the specified note, not even if its the last node left in the cluster!
To display an overview of all currently applying resource constraint scores:
[root@centos1 ~]# crm_simulate -sL
Current cluster status:
Online: [ centos1.localdomain centos2.localdomain centos3.localdomain ]
fence_centos1 (stonith:fence_xvm): Started centos3.localdomain
fence_centos2 (stonith:fence_xvm): Started centos3.localdomain
fence_centos3 (stonith:fence_xvm): Started centos3.localdomain
pcmk__native_allocate: fence_centos1 allocation score on centos1.localdomain: 0
pcmk__native_allocate: fence_centos1 allocation score on centos2.localdomain: 0
pcmk__native_allocate: fence_centos1 allocation score on centos3.localdomain: 0
pcmk__native_allocate: fence_centos2 allocation score on centos1.localdomain: -INFINITY
pcmk__native_allocate: fence_centos2 allocation score on centos2.localdomain: -INFINITY
pcmk__native_allocate: fence_centos2 allocation score on centos3.localdomain: 0
pcmk__native_allocate: fence_centos3 allocation score on centos1.localdomain: -INFINITY
pcmk__native_allocate: fence_centos3 allocation score on centos2.localdomain: -INFINITY
pcmk__native_allocate: fence_centos3 allocation score on centos3.localdomain: 0