def health_checks(self):
""" perform checks on elements that affect the reported state of the cluster """
# The idea here is to perform the most significant checks first
# so the message list appears in a priority order
# 1. Check the volumes
for volume_name in self.volume:
this_volume = self.volume[volume_name]
if 'down' in this_volume.volume_state:
self.messages.append("Volume '%s' is down" % volume_name)
self.status = 'unhealthy'
if 'partial' in this_volume.volume_state:
self.messages.append(
"Volume '%s' is in a PARTIAL state, some data is inaccessible data, due to missing bricks"
% volume_name)
self.messages.append("WARNING -> Write requests may fail against volume '%s'" % this_volume.name)
self.status = 'unhealthy'
# 2. Check for conditions detected at the node level
for uuid in self.node:
this_node = self.node[uuid]
if this_node.state != '1':
# https://bugzilla.redhat.com/show_bug.cgi?id=1254514,
# node_name comes as empty
# self.messages.append("Cluster node '%s' is down" %
# (this_node.node_name()))
self.nodes_down += 1
self.status = 'unhealthy'
if this_node.self_heal_enabled != this_node.self_heal_active:
# https://bugzilla.redhat.com/show_bug.cgi?id=1254514,
# decided to remove the self-heal status as it is
# redunant information
# self.messages.append("Self heal daemon is down on %s" % (this_node.node_name()))
self.status = 'unhealthy'
# Print the number of nodes that are down
if self.nodes_down == 1:
self.messages.append("One of the nodes in the cluster is down")
elif self.nodes_down > 1:
self.messages.append("%s nodes in the cluster are down"%self.nodes_down)
# 3. Check the bricks
for brick_name in self.brick:
this_brick = self.brick[brick_name]
# 3.1 check for state
if not this_brick.up:
self.messages.append(
"Brick %s in volume '%s' is down/unavailable" % (brick_name, this_brick.owning_volume))
# 3.2 check for best practice goes here (minor error messages - FUTURE)
if self.bricks_active < Brick.brick_count():
self.messages.append("INFO -> Not all bricks are online, so capacity provided is NOT accurate")
def health_checks(self):
""" perform checks on elements that affect the reported state of the cluster """
# The idea here is to perform the most significant checks first
# so the message list appears in a priority order
# 1. Check the volumes
for volume_name in self.volume:
this_volume = self.volume[volume_name]
if 'down' in this_volume.volume_state:
self.messages.append("Volume '%s' is down" % volume_name)
self.status = 'unhealthy'
if 'partial' in this_volume.volume_state:
self.messages.append(
"Volume '%s' is in a PARTIAL state, some data is inaccessible data, due to missing bricks"
% volume_name)
self.messages.append("WARNING -> Write requests may fail against volume '%s'" % this_volume.name)
self.status = 'unhealthy'
# 2. Check for conditions detected at the node level
for uuid in self.node:
this_node = self.node[uuid]
if this_node.state != '1':
# https://bugzilla.redhat.com/show_bug.cgi?id=1254514,
# node_name comes as empty
# self.messages.append("Cluster node '%s' is down" %
# (this_node.node_name()))
self.nodes_down += 1
self.status = 'unhealthy'
if this_node.self_heal_enabled != this_node.self_heal_active:
# https://bugzilla.redhat.com/show_bug.cgi?id=1254514,
# decided to remove the self-heal status as it is
# redunant information
# self.messages.append("Self heal daemon is down on %s" % (this_node.node_name()))
self.status = 'unhealthy'
# Print the number of nodes that are down
if self.nodes_down == 1:
self.messages.append("One of the nodes in the cluster is down")
elif self.nodes_down > 1:
self.messages.append("%s nodes in the cluster are down"%self.nodes_down)
# 3. Check the bricks
for brick_name in self.brick:
this_brick = self.brick[brick_name]
# 3.1 check for state
if not this_brick.up:
self.messages.append(
"Brick %s in volume '%s' is down/unavailable" % (brick_name, this_brick.owning_volume))
# 3.2 check for best practice goes here (minor error messages - FUTURE)
if self.bricks_active < Brick.brick_count():
self.messages.append("INFO -> Not all bricks are online, so capacity provided is NOT accurate")
def health_checks(self):
""" perform checks on elements that affect the reported state of the cluster """
# The idea here is to perform the most significant checks first
# so the message list appears in a priority order
# 1. Check the volumes
for volume_name in self.volume:
this_volume = self.volume[volume_name]
if 'down' in this_volume.volume_state:
self.messages.append("Volume '%s' is down" % volume_name)
self.status = 'unhealthy'
if 'partial' in this_volume.volume_state:
self.messages.append(
"Volume '%s' is in a PARTIAL state, some data is inaccessible data, due to missing bricks"
% volume_name)
self.messages.append("WARNING -> Write requests may fail against volume '%s'" % this_volume.name)
self.status = 'unhealthy'
# 2. Check for conditions detected at the node level
for uuid in self.node:
this_node = self.node[uuid]
if this_node.state != '1':
self.messages.append("Cluster node '%s' is down" % (this_node.node_name()))
self.status = 'unhealthy'
if this_node.self_heal_enabled != this_node.self_heal_active:
self.messages.append("Self heal daemon is down on %s" % (this_node.node_name()))
self.status = 'unhealthy'
# 3. Check the bricks
for brick_name in self.brick:
this_brick = self.brick[brick_name]
# 3.1 check for state
if not this_brick.up:
self.messages.append(
"Brick %s in volume '%s' is down/unavailable" % (brick_name, this_brick.owning_volume))
# 3.2 check for best practice goes here (minor error messages - FUTURE)
if self.bricks_active < Brick.brick_count():
self.messages.append("INFO -> Not all bricks are online, so capacity provided is NOT accurate")
def define_volumes(self):
""" Create the volume + brick objects """
if self.output_mode == 'console' and not cfg.no_progress_msgs:
# print a progress message
sys.stdout.write("Building volume objects" + " " * 20 + "\n\r\x1b[A")
cmd = GlusterCommand("gluster vol info --xml")
cmd.run()
# (rc, vol_info) = issueCMD("gluster vol info --xml")
xml_string = ''.join(cmd.stdout)
xml_root = ETree.fromstring(xml_string)
vol_elements = xml_root.findall('.//volume')
for vol_object in vol_elements:
# set up a dict for the initial definition of the volume
vol_dict = {}
# build a dict for the initial volume settings. An attribute error results in a default
# value being assigned (e.g. on older glusterfs disperse related fields are missing)
for attr in Volume.volume_attr:
try:
vol_dict[attr] = vol_object.find('./' + attr).text
except AttributeError:
vol_dict[attr] = '0'
# create a volume object, for this volume
new_volume = Volume(vol_dict)
self.volume[new_volume.name] = new_volume
if cfg.debug:
print "defineVolumes. Adding volume %s" % new_volume.name
# add information about any volume options
opt_nodes = vol_object.findall('.//option')
for option in opt_nodes:
for n in option.getchildren():
if n.tag == 'name':
key = n.text
elif n.tag == 'value':
value = n.text
new_volume.options[key] = value
# Protocols are enabled by default, so we look
# for the volume tuning options that turn them
# off
if key == 'user.cifs':
if value in ['disable', 'off', 'false']:
new_volume.protocol['SMB'] = 'off'
elif key == 'nfs.disable':
if value in ['on', 'true']:
new_volume.protocol['NFS'] = 'off'
# get bricks listed against this volume, and create the Brick object(s)
brick_nodes = vol_object.findall('.//brick')
# list holding brick paths
repl = []
ctr = 1
for brick in brick_nodes:
brick_path = brick.text
new_volume.brick_order.append(brick_path)
(hostname, pathname) = brick_path.split(':')
if cfg.debug:
print "defineVolumes. Adding brick %s to %s" % (brick_path,
new_volume.name)
node_uuid = self.get_node(hostname)
# add this bricks owning node to the volume's attributes
try:
new_volume.node[node_uuid] = self.node[node_uuid]
except KeyError:
print "Unable to associate brick %s with a peer in the cluster, possibly due" % brick_path
print "to name lookup failures. If the nodes are not registered (fwd & rev)"
print "to dns, add local entries for your cluster nodes in the the /etc/hosts file"
sys.exit(16)
new_brick = Brick(brick_path, self.node[node_uuid], new_volume.name)
# Add the brick to the cluster and volume
self.brick[brick_path] = new_brick
new_volume.brick[brick_path] = new_brick
# add this brick to the owning node
brick_owner = self.node[node_uuid]
brick_owner.brick[brick_path] = new_brick
if (new_volume.replicaCount > 1) or (new_volume.disperseCount > 0):
repl.append(brick_path)
bricks_per_subvolume = max(new_volume.replicaCount, new_volume.disperseCount)
ctr += 1
if ctr > bricks_per_subvolume:
ctr = 1
# add this replica set to the volume's info
new_volume.subvolumes.append(repl)
# drop all elements from temporary list
repl = []
# By default from gluster 3.3 onwards, self heal is enabled for
# all replicated/disperse volumes. We look at the volume type, and if it
# is replicated and hasn't had self-heal explicitly disabled the
# self heal state is inferred against the nodes that contain the
# bricks for the volume. With this state in place, the updateState
# method can cross-check to see what is actually happening
if ('replicate' in new_volume.typeStr.lower()) or ('disperse' in new_volume.typeStr.lower()):
heal_enabled = True # assume it's on
if 'cluster.self-heal-daemon' in new_volume.options:
if new_volume.options['cluster.self-heal-daemon'].lower() in ['off', 'false']:
heal_enabled = False
new_volume.self_heal_enabled = heal_enabled
if heal_enabled:
node_set = set() # use a set to maintain a unique group of nodes
for brick_path in new_volume.brick:
this_brick = self.brick[brick_path]
this_node = this_brick.node
node_set.add(this_node)
this_node.self_heal_enabled = True
self.sh_enabled = len(node_set)
self.volume_count = Volume.volume_count()
self.brick_count = Brick.brick_count()
def define_volumes(self):
""" Create the volume + brick objects """
if self.output_mode == 'console' and not cfg.no_progress_msgs:
# print a progress message
sys.stdout.write("Building volume objects" + " " * 20 + "\n\r\x1b[A")
cmd = GlusterCommand("gluster vol info --xml", timeout=cfg.CMD_TIMEOUT)
cmd.run()
# (rc, vol_info) = issueCMD("gluster vol info --xml")
xml_string = ''.join(cmd.stdout)
xml_root = ETree.fromstring(xml_string)
vol_elements = xml_root.findall('.//volume')
for vol_object in vol_elements:
# set up a dict for the initial definition of the volume
vol_dict = {}
# build a dict for the initial volume settings. An attribute error results in a default
# value being assigned (e.g. on older glusterfs disperse related fields are missing)
for attr in Volume.volume_attr:
try:
vol_dict[attr] = vol_object.find('./' + attr).text
except AttributeError:
vol_dict[attr] = '0'
# create a volume object, for this volume
new_volume = Volume(vol_dict)
self.volume[new_volume.name] = new_volume
if cfg.debug:
print "defineVolumes. Adding volume %s" % new_volume.name
# add information about any volume options
opt_nodes = vol_object.findall('.//option')
for option in opt_nodes:
for n in option.getchildren():
if n.tag == 'name':
key = n.text
elif n.tag == 'value':
value = n.text
new_volume.options[key] = value
# Protocols are enabled by default, so we look
# for the volume tuning options that turn them
# off
if key == 'user.cifs':
if value in ['disable', 'off', 'false']:
new_volume.protocol['SMB'] = 'off'
elif key == 'nfs.disable':
if value in ['on', 'true']:
new_volume.protocol['NFS'] = 'off'
# get bricks listed against this volume, and create the Brick object(s)
brick_nodes = vol_object.findall('.//brick')
# list holding brick paths
repl = []
ctr = 1
for brick in brick_nodes:
brick_path = brick.text
new_volume.brick_order.append(brick_path)
(hostname, pathname) = brick_path.split(':')
if cfg.debug:
print "defineVolumes. Adding brick %s to %s" % (brick_path,
new_volume.name)
node_uuid = self.get_node(hostname)
# add this bricks owning node to the volume's attributes
try:
new_volume.node[node_uuid] = self.node[node_uuid]
except KeyError:
print "Unable to associate brick %s with a peer in the cluster, possibly due" % brick_path
print "to name lookup failures. If the nodes are not registered (fwd & rev)"
print "to dns, add local entries for your cluster nodes in the the /etc/hosts file"
sys.exit(16)
new_brick = Brick(brick_path, self.node[node_uuid], new_volume.name)
# Add the brick to the cluster and volume
self.brick[brick_path] = new_brick
new_volume.brick[brick_path] = new_brick
# add this brick to the owning node
brick_owner = self.node[node_uuid]
brick_owner.brick[brick_path] = new_brick
if (new_volume.replicaCount > 1) or (new_volume.disperseCount > 0):
repl.append(brick_path)
bricks_per_subvolume = max(new_volume.replicaCount, new_volume.disperseCount)
ctr += 1
if ctr > bricks_per_subvolume:
ctr = 1
# add this replica set to the volume's info
new_volume.subvolumes.append(repl)
# drop all elements from temporary list
repl = []
# By default from gluster 3.3 onwards, self heal is enabled for
# all replicated/disperse volumes. We look at the volume type, and if it
# is replicated and hasn't had self-heal explicitly disabled the
# self heal state is inferred against the nodes that contain the
# bricks for the volume. With this state in place, the updateState
# method can cross-check to see what is actually happening
if ('replicate' in new_volume.typeStr.lower()) or ('disperse' in new_volume.typeStr.lower()):
heal_enabled = True # assume it's on
if 'cluster.self-heal-daemon' in new_volume.options:
if new_volume.options['cluster.self-heal-daemon'].lower() in ['off', 'false']:
heal_enabled = False
new_volume.self_heal_enabled = heal_enabled
if heal_enabled:
node_set = set() # use a set to maintain a unique group of nodes
for brick_path in new_volume.brick:
this_brick = self.brick[brick_path]
this_node = this_brick.node
node_set.add(this_node)
this_node.self_heal_enabled = True
self.sh_enabled = len(node_set)
self.volume_count = Volume.volume_count()
self.brick_count = Brick.brick_count()
