E
E
Esergion2014-04-08 13:49:10
Clusters
Esergion, 2014-04-08 13:49:10

Corosync, pacemaker. How to remove monitoring of some resources from certain cluster nodes?

Good time!
Interested in the possibility of launching cluster resources on only two nodes out of the set.
For example, only the PGSERVER group, which includes the drbd, FS, postgresql, DB_virtual_IP resources, should jump between DB1 and DB2.
And between WEB1 and WEB2 nodes only WEB_virtual_IP.
In principle, everything has already been done, but

Online: [ db1 db2 api2 api1 ]

 Master/Slave Set: ms_drbd_pg [drbd_pg]
     Masters: [ db1 ]
     Slaves: [ db2 ]
 Resource Group: PGServer
     pg_fs      (ocf::heartbeat:Filesystem):    Started db1
     pg_lsb     (lsb:postgresql):       Started db1
     pg_vip     (ocf::heartbeat:IPaddr2):       Started db1
api_vip (ocf::heartbeat:IPaddr2):       Started api1

Failed actions:
    drbd_pg:0_monitor_0 (node=api2, call=2, rc=5, status=complete): not installed
    pg_lsb_monitor_0 (node=api2, call=4, rc=5, status=complete): not installed
    drbd_pg:0_monitor_0 (node=api1, call=2, rc=5, status=complete): not installed
    pg_lsb_monitor_0 (node=api1, call=4, rc=5, status=complete): not installed

But why is he trying to check these resources on a node that is not intended for these resources?
Here is a piece of the config, which describes which nodes resources should work on.
location cli-prefer-PGServer PGServer \
        rule $id="cli-prefer-rule-PGServer" inf: #uname eq db1
location cli-prefer-api_vip api_vip \
        rule $id="cli-prefer-rule-api_vip" inf: #uname eq api1
location loc-drbd_pg_01 ms_drbd_pg -inf: api1
location loc-drbd_pg_02 ms_drbd_pg -inf: api2
location loc-pgserver_01 PGServer -inf: api1
location loc-pgserver_02 PGServer -inf: api2
location loc_api_vip_01 api_vip 100: api1
location loc_api_vip_02 api_vip 10: api2
location loc_api_vip_03 api_vip -inf: db1
location loc_api_vip_04 api_vip -inf: db2

How to make it so that there are no "Failed actions" in crm_mon, i.e. so that the cluster does not even try to monitor some resources on some nodes?

Answer the question

In order to leave comments, you need to log in

Didn't find what you were looking for?

Ask your question

Ask a Question

731 491 924 answers to any question