def unconfigure_target_ha(primary, ha_label, uuid):
'''
Unconfigure the target high availability
Return: Value using simple return protocol
'''
with PreservePacemakerCorosyncState():
if get_resource_location(ha_label):
return agent_error("cannot unconfigure-ha: %s is still running " %
ha_label)
if primary:
result = cibadmin(
["-D", "-X",
"<rsc_location id=\"%s-primary\">" % ha_label])
result = cibadmin(["-D", "-X", "<primitive id=\"%s\">" % ha_label])
if result.rc != 0 and result.rc != 234:
return agent_error("Error %s trying to cleanup resource %s" %
(result.rc, ha_label))
else:
result = cibadmin(
["-D", "-X",
"<rsc_location id=\"%s-secondary\">" % ha_label])
return agent_result_ok
def _do_configure_pacemaker(pc):
# ignoring quorum should only be done on clusters of 2
if len(pc.nodes) > 2:
no_quorum_policy = "stop"
else:
no_quorum_policy = "ignore"
error = _unconfigure_fencing()
if error:
return error
# this could race with other cluster members to make sure
# any errors are only due to it already existing
try:
cibadmin([
"--create",
"-o",
"resources",
"-X",
'<primitive class="stonith" id="st-fencing" type="fence_chroma"/>',
])
except Exception as e:
rc, stdout, stderr = AgentShell.run_old(
["crm_resource", "--locate", "--resource", "st-fencing"])
if rc == 0: # no need to do the rest if another member is already doing it
return None
else:
return e.message
pc.create_update_properyset(
"cib-bootstrap-options",
{
"no-quorum-policy": no_quorum_policy,
"symmetric-cluster": "true",
"cluster-infrastructure": "openais",
"stonith-enabled": "true",
},
)
def set_rsc_default(name, value):
"""
:param name: attribute to set
:param value: value to set
:return: None if an error else a canned error message
"""
return AgentShell.run_canned_error_message([
"crm_attribute",
"--type",
"rsc_defaults",
"--attr-name",
name,
"--attr-value",
value,
])
return (set_rsc_default("resource-stickiness", "1000")
or set_rsc_default("failure-timeout", RSRC_FAIL_WINDOW) or
set_rsc_default("migration-threshold", RSRC_FAIL_MIGRATION_COUNT))
def _unconfigure_fencing():
try:
cibadmin([
"--delete", "-o", "resources", "-X",
"<primitive class=\"stonith\" id=\"st-fencing\" type=\"fence_chroma\"/>"
])
return None
except Exception as e:
return e.message
def unconfigure_pacemaker():
# only unconfigure if we are the only node in the cluster
# but first, see if pacemaker is up to answer this
if not _pacemaker_running():
# and just skip doing this if it's not
return agent_result_ok
if _get_cluster_size() < 2:
# last node, nuke the CIB
cibadmin(["-f", "-E"])
return agent_ok_or_error(pacemaker_service.stop()
or pacemaker_service.disable())
def _unconfigure_fencing():
try:
cibadmin([
"--delete",
"-o",
"resources",
"-X",
'<primitive class="stonith" id="st-fencing" type="fence_chroma"/>',
])
return None
except Exception as e:
return e.message
def delete_node(nodename):
rc, stdout, stderr = AgentShell.run_old(['crm_node', '-l'])
node_id = None
for line in stdout.split('\n'):
node_id, name, status = line.split(" ")
if name == nodename:
break
AgentShell.try_run(['crm_node', '--force', '-R', node_id])
cibadmin(
["--delete", "-o", "nodes", "-X",
"<node uname=\"%s\"/>" % nodename])
cibadmin([
"--delete", "-o", "nodes", "--crm_xml",
"<node_state uname=\"%s\"/>" % nodename
])
def configure_target_ha(primary, device, ha_label, uuid, mount_point):
'''
Configure the target high availability
Return: Value using simple return protocol
'''
if primary:
# If the target already exists with the same params, skip.
# If it already exists with different params, that is an error
rc, stdout, stderr = AgentShell.run_old(
["crm_resource", "-r", ha_label, "-g", "target"])
if rc == 0:
info = _get_target_config(stdout.rstrip("\n"))
if info['bdev'] == device and info['mntpt'] == mount_point:
return agent_result_ok
else:
return agent_error(
"A resource with the name %s already exists" % ha_label)
tmp_f, tmp_name = tempfile.mkstemp()
os.write(
tmp_f,
"<primitive class=\"ocf\" provider=\"chroma\" type=\"Target\" id=\"%s\">\
<meta_attributes id=\"%s-meta_attributes\">\
<nvpair name=\"target-role\" id=\"%s-meta_attributes-target-role\" value=\"Stopped\"/>\
</meta_attributes>\
<operations id=\"%s-operations\">\
<op id=\"%s-monitor-5\" interval=\"5\" name=\"monitor\" timeout=\"60\"/>\
<op id=\"%s-start-0\" interval=\"0\" name=\"start\" timeout=\"300\"/>\
<op id=\"%s-stop-0\" interval=\"0\" name=\"stop\" timeout=\"300\"/>\
</operations>\
<instance_attributes id=\"%s-instance_attributes\">\
<nvpair id=\"%s-instance_attributes-target\" name=\"target\" value=\"%s\"/>\
</instance_attributes>\
</primitive>" % (ha_label, ha_label, ha_label, ha_label, ha_label, ha_label,
ha_label, ha_label, ha_label, uuid))
os.close(tmp_f)
cibadmin(["-o", "resources", "-C", "-x", "%s" % tmp_name])
score = 20
preference = "primary"
else:
score = 10
preference = "secondary"
# Hostname. This is a shorterm point fix that will allow us to make HP2 release more functional. Between el6 and el7
# (truthfully we should probably be looking at Pacemaker or Corosync versions) Pacemaker started to use fully qualified
# domain names rather than just the nodename. lotus-33vm15.lotus.hpdd.lab.intel.com vs lotus-33vm15. To keep compatiblity
# easily we have to make the contraints follow the same fqdn vs node.
if platform_info.distro_version >= 7.0:
node = socket.getfqdn()
else:
node = os.uname()[1]
result = cibadmin([
"-o", "constraints", "-C", "-X",
"<rsc_location id=\"%s-%s\" node=\"%s\" rsc=\"%s\" score=\"%s\"/>" %
(ha_label, preference, node, ha_label, score)
])
if result.rc == 76:
return agent_error("A constraint with the name %s-%s already exists" %
(ha_label, preference))
_mkdir_p_concurrent(mount_point)
return agent_result_ok
