mgmtd cannot handle templated primitives any more

[fgtham]

When the CIB contains primitives based on templates, crm_gui cannot connect to the cluster's mgmtd any longer, as mgmtd dies in the process. The cluster can no longer be managed via the GUI. CLI management is not affected.

Software used:

• corosync 1.3.0
• pacemaker 1.1.12-561c4cf
• pacemaker-mgmt 79579ac

Cluster configuration (fresh single-node pacemaker install with STONITH disabled):

# crm configure show
node desk
property cib-bootstrap-options: \
        dc-version=1.1.12-561c4cf \
        cluster-infrastructure="classic openais (with plugin)" \
        expected-quorum-votes=2 \
        stonith-enabled=false

Steps to reproduce:

1. Connect crm_gui to cluster, verify connection is ok

2. Create template:

   # crm configure rsc_template t_ip ocf:heartbeat:IPaddr2 \
   params nic=br0 cidr_netmask=24 \
   operations \$id=t_ip-ops \
   op monitor interval=10s timeout=20s
   

3. Create primitive from template:

   # crm configure primitive p_gw @t_ip params ip=192.168.123.123
   

Result:

After resource creation mgmtd dies, crm_gui is no longer connected:

Jan 26 12:40:33 desk mgmtd[1723]: [1723]: debug: recv msg: cib_query
                                  cib
Jan 26 12:40:33 desk mgmtd[1723]: [1723]: debug: set_crm: client_id=1 cib_conn=0x16ec680 cib_name=(null)
Jan 26 12:40:33 desk mgmtd[1723]: [1723]: info: CIB query: cib
Jan 26 12:40:33 desk mgmtd[1723]: [1723]: debug: send msg: o
                                  <cib crm_feature_set="3.0.9" validate-with="pacemaker-2.0" epoch="60" num_updates="1"
 admin_epoch="0" cib-last-written="Mon Jan 26 12:40:32 2015" have-quorum="0" dc-uuid="desk">
 ...
                                  </cib>
Jan 26 12:40:33 desk mgmtd[1723]: [1723]: debug: recv msg: active_cib
Jan 26 12:40:33 desk mgmtd[1723]: [1723]: debug: set_crm: client_id=1 cib_conn=0x16ec680 cib_name=(null)
Jan 26 12:40:33 desk mgmtd[1723]: [1723]: debug: send msg: o
Jan 26 12:40:33 desk mgmtd[1723]: [1723]: debug: recv msg: all_nodes
Jan 26 12:40:33 desk mgmtd[1723]: [1723]: debug: set_crm: client_id=1 cib_conn=0x16ec680 cib_name=(null)
Jan 26 12:40:33 desk mgmtd[1723]: [1723]: debug: send msg: f
Jan 26 12:40:33 desk mgmtd[1723]: [1723]: debug: recv msg: crm_nodes
Jan 26 12:40:33 desk mgmtd[1723]: [1723]: debug: set_crm: client_id=1 cib_conn=0x16ec680 cib_name=(null)
Jan 26 12:40:33 desk pacemakerd[1383]: notice: pcmk_child_exit: Child process mgmtd terminated with signal 6 (pid=1723,
 core=0)
Jan 26 12:40:33 desk corosync[1382]: [CPG   ] got mcast request on 0x14cd100
Jan 26 12:40:33 desk pacemakerd[1383]: notice: pcmk_process_exit: Respawning failed child process: mgmtd

Subsequent connection attempts fail until the resource is deleted.