[gnuoy,r=wolsen] Use current DC to determine leader.

hooks/charmhelpers/contrib/hahelpers/cluster.py

@@ -52,6 +52,8 @@ from charmhelpers.core.strutils import (
     bool_from_string,
 )
 
+DC_RESOURCE_NAME = 'DC'
+
 
 class HAIncompleteConfig(Exception):
     pass
@@ -95,6 +97,27 @@ def is_clustered():
     return False
 
 
+def is_crm_dc():
+    """
+    Determine leadership by querying the pacemaker Designated Controller
+    """
+    cmd = ['crm', 'status']
+    try:
+        status = subprocess.check_output(cmd, stderr=subprocess.STDOUT)
+        if not isinstance(status, six.text_type):
+            status = six.text_type(status, "utf-8")
+    except subprocess.CalledProcessError:
+        return False
+    current_dc = ''
+    for line in status.split('\n'):
+        if line.startswith('Current DC'):
+            # Current DC: juju-lytrusty-machine-2 (168108163) - partition with quorum
+            current_dc = line.split(':')[1].split()[0]
+    if current_dc == get_unit_hostname():
+        return True
+    return False
+
+
 @retry_on_exception(5, base_delay=2, exc_type=CRMResourceNotFound)
 def is_crm_leader(resource, retry=False):
     """
@@ -104,6 +127,8 @@ def is_crm_leader(resource, retry=False):
     We allow this operation to be retried to avoid the possibility of getting a
     false negative. See LP #1396246 for more info.
     """
+    if resource == DC_RESOURCE_NAME:
+        return is_crm_dc()
     cmd = ['crm', 'resource', 'show', resource]
     try:
         status = subprocess.check_output(cmd, stderr=subprocess.STDOUT)

hooks/charmhelpers/contrib/peerstorage/__init__.py

@@ -73,6 +73,8 @@ def peer_retrieve_by_prefix(prefix, relation_name='cluster', delimiter='_',
     exc_list = exc_list if exc_list else []
     peerdb_settings = peer_retrieve('-', relation_name=relation_name)
     matched = {}
+    if peerdb_settings is None:
+        return matched
     for k, v in peerdb_settings.items():
         full_prefix = prefix + delimiter
         if k.startswith(full_prefix):

hooks/charmhelpers/core/hookenv.py

@@ -21,12 +21,14 @@
 #  Charm Helpers Developers <juju@lists.ubuntu.com>
 
 from __future__ import print_function
+from functools import wraps
 import os
 import json
 import yaml
 import subprocess
 import sys
 import errno
+import tempfile
 from subprocess import CalledProcessError
 
 import six
@@ -58,15 +60,17 @@ def cached(func):
 
     will cache the result of unit_get + 'test' for future calls.
     """
+    @wraps(func)
     def wrapper(*args, **kwargs):
         global cache
         key = str((func, args, kwargs))
         try:
             return cache[key]
         except KeyError:
-            res = func(*args, **kwargs)
-            cache[key] = res
-            return res
+            pass  # Drop out of the exception handler scope.
+        res = func(*args, **kwargs)
+        cache[key] = res
+        return res
     return wrapper
 
 
@@ -178,7 +182,7 @@ def local_unit():
 
 def remote_unit():
     """The remote unit for the current relation hook"""
-    return os.environ['JUJU_REMOTE_UNIT']
+    return os.environ.get('JUJU_REMOTE_UNIT', None)
 
 
 def service_name():
@@ -250,6 +254,12 @@ class Config(dict):
         except KeyError:
             return (self._prev_dict or {})[key]
 
+    def get(self, key, default=None):
+        try:
+            return self[key]
+        except KeyError:
+            return default
+
     def keys(self):
         prev_keys = []
         if self._prev_dict is not None:
@@ -353,14 +363,29 @@ def relation_set(relation_id=None, relation_settings=None, **kwargs):
     """Set relation information for the current unit"""
     relation_settings = relation_settings if relation_settings else {}
     relation_cmd_line = ['relation-set']
+    accepts_file = "--file" in subprocess.check_output(
+        relation_cmd_line + ["--help"])
     if relation_id is not None:
         relation_cmd_line.extend(('-r', relation_id))
-    for k, v in (list(relation_settings.items()) + list(kwargs.items())):
-        if v is None:
-            relation_cmd_line.append('{}='.format(k))
-        else:
-            relation_cmd_line.append('{}={}'.format(k, v))
-    subprocess.check_call(relation_cmd_line)
+    settings = relation_settings.copy()
+    settings.update(kwargs)
+    if accepts_file:
+        # --file was introduced in Juju 1.23.2. Use it by default if
+        # available, since otherwise we'll break if the relation data is
+        # too big. Ideally we should tell relation-set to read the data from
+        # stdin, but that feature is broken in 1.23.2: Bug #1454678.
+        with tempfile.NamedTemporaryFile(delete=False) as settings_file:
+            settings_file.write(yaml.safe_dump(settings).encode("utf-8"))
+        subprocess.check_call(
+            relation_cmd_line + ["--file", settings_file.name])
+        os.remove(settings_file.name)
+    else:
+        for key, value in settings.items():
+            if value is None:
+                relation_cmd_line.append('{}='.format(key))
+            else:
+                relation_cmd_line.append('{}={}'.format(key, value))
+        subprocess.check_call(relation_cmd_line)
     # Flush cache of any relation-gets for local unit
     flush(local_unit())
 
@@ -509,6 +534,11 @@ def unit_get(attribute):
         return None
 
 
+def unit_public_ip():
+    """Get this unit's public IP address"""
+    return unit_get('public-address')
+
+
 def unit_private_ip():
     """Get this unit's private IP address"""
     return unit_get('private-address')
@@ -605,3 +635,49 @@ def action_fail(message):
 
     The results set by action_set are preserved."""
     subprocess.check_call(['action-fail', message])
+
+
+def status_set(workload_state, message):
+    """Set the workload state with a message
+
+    Use status-set to set the workload state with a message which is visible
+    to the user via juju status. If the status-set command is not found then
+    assume this is juju < 1.23 and juju-log the message unstead.
+
+    workload_state -- valid juju workload state.
+    message        -- status update message
+    """
+    valid_states = ['maintenance', 'blocked', 'waiting', 'active']
+    if workload_state not in valid_states:
+        raise ValueError(
+            '{!r} is not a valid workload state'.format(workload_state)
+        )
+    cmd = ['status-set', workload_state, message]
+    try:
+        ret = subprocess.call(cmd)
+        if ret == 0:
+            return
+    except OSError as e:
+        if e.errno != errno.ENOENT:
+            raise
+    log_message = 'status-set failed: {} {}'.format(workload_state,
+                                                    message)
+    log(log_message, level='INFO')
+
+
+def status_get():
+    """Retrieve the previously set juju workload state
+
+    If the status-set command is not found then assume this is juju < 1.23 and
+    return 'unknown'
+    """
+    cmd = ['status-get']
+    try:
+        raw_status = subprocess.check_output(cmd, universal_newlines=True)
+        status = raw_status.rstrip()
+        return status
+    except OSError as e:
+        if e.errno == errno.ENOENT:
+            return 'unknown'
+        else:
+            raise

hooks/charmhelpers/core/host.py

@@ -90,7 +90,7 @@ def service_available(service_name):
             ['service', service_name, 'status'],
             stderr=subprocess.STDOUT).decode('UTF-8')
     except subprocess.CalledProcessError as e:
-        return 'unrecognized service' not in e.output
+        return b'unrecognized service' not in e.output
     else:
         return True
 

hooks/charmhelpers/core/services/base.py

@@ -17,7 +17,7 @@
 import os
 import re
 import json
-from collections import Iterable
+from collections import Iterable, OrderedDict
 
 from charmhelpers.core import host
 from charmhelpers.core import hookenv
@@ -119,7 +119,7 @@ class ServiceManager(object):
         """
         self._ready_file = os.path.join(hookenv.charm_dir(), 'READY-SERVICES.json')
         self._ready = None
-        self.services = {}
+        self.services = OrderedDict()
         for service in services or []:
             service_name = service['service']
             self.services[service_name] = service

hooks/charmhelpers/fetch/__init__.py

@@ -158,7 +158,7 @@ def filter_installed_packages(packages):
 
 def apt_cache(in_memory=True):
     """Build and return an apt cache"""
-    import apt_pkg
+    from apt import apt_pkg
     apt_pkg.init()
     if in_memory:
         apt_pkg.config.set("Dir::Cache::pkgcache", "")

hooks/percona_hooks.py

@@ -56,6 +56,7 @@ from charmhelpers.contrib.database.mysql import (
     PerconaClusterHelper,
 )
 from charmhelpers.contrib.hahelpers.cluster import (
+    DC_RESOURCE_NAME,
     peer_units,
     oldest_peer,
     eligible_leader,
@@ -74,7 +75,6 @@ from charmhelpers.contrib.charmsupport import nrpe
 
 hooks = Hooks()
 
-LEADER_RES = 'grp_percona_cluster'
 RES_MONITOR_PARAMS = ('params user="sstuser" password="%(sstpass)s" '
                       'pid="/var/run/mysqld/mysqld.pid" '
                       'socket="/var/run/mysqld/mysqld.sock" '
@@ -207,7 +207,7 @@ def cluster_changed():
 @hooks.hook('db-relation-changed')
 @hooks.hook('db-admin-relation-changed')
 def db_changed(relation_id=None, unit=None, admin=None):
-    if not eligible_leader(LEADER_RES):
+    if not eligible_leader(DC_RESOURCE_NAME):
         log('Service is peered, clearing db relation'
             ' as this service unit is not the leader')
         relation_clear(relation_id)
@@ -269,7 +269,7 @@ def configure_db_for_hosts(hosts, database, username, db_helper):
 # TODO: This could be a hook common between mysql and percona-cluster
 @hooks.hook('shared-db-relation-changed')
 def shared_db_changed(relation_id=None, unit=None):
-    if not eligible_leader(LEADER_RES):
+    if not eligible_leader(DC_RESOURCE_NAME):
         relation_clear(relation_id)
         # Each unit needs to set the db information otherwise if the unit
         # with the info dies the settings die with it Bug# 1355848
@@ -412,7 +412,7 @@ def ha_relation_joined():
                                             password=cfg_passwd)
     resource_params = {'res_mysql_vip': vip_params,
                        'res_mysql_monitor':
-                           RES_MONITOR_PARAMS % {'sstpass': sstpsswd}}
+                       RES_MONITOR_PARAMS % {'sstpass': sstpsswd}}
     groups = {'grp_percona_cluster': 'res_mysql_vip'}
 
     clones = {'cl_mysql_monitor': 'res_mysql_monitor meta interleave=true'}
@@ -437,7 +437,7 @@ def ha_relation_joined():
 @hooks.hook('ha-relation-changed')
 def ha_relation_changed():
     clustered = relation_get('clustered')
-    if (clustered and is_leader(LEADER_RES)):
+    if (clustered and is_leader(DC_RESOURCE_NAME)):
         log('Cluster configured, notifying other services')
         # Tell all related services to start using the VIP
         for r_id in relation_ids('shared-db'):