Bump version to 0.9.25

Just a minor quibble and not really impactful.

.version

@@ -1 +1 @@
-0.9.23
+0.9.25

README.md

@@ -42,6 +42,22 @@ To get started with PVC, please see the [About](https://parallelvirtualcluster.r
 
 ## Changelog
 
+#### v0.9.25
+
+  * [Node Daemon] Returns to Rados library calls for Ceph due to performance problems
+  * [Node Daemon] Adds a date output to keepalive messages
+  * [Daemons] Configures ZK connection logging only for persistent connections
+  * [API Provisioner] Add context manager-based chroot to Debootstrap example script
+  * [Node Daemon] Fixes a bug where shutdown daemon state was overwritten
+
+#### v0.9.24
+
+  * [Node Daemon] Removes Rados module polling of Ceph cluster and returns to command-based polling for timeout purposes, and removes some flaky return statements
+  * [Node Daemon] Removes flaky Zookeeper connection renewals that caused problems
+  * [CLI Client] Allow raw lists of clusters from `pvc cluster list`
+  * [API Daemon] Fixes several issues when getting VM data without stats
+  * [API Daemon] Fixes issues with removing VMs while disks are still in use (failed provisioning, etc.)
+
 #### v0.9.23
 
   * [Daemons] Fixes a critical overwriting bug in zkhandler when schema paths are not yet valid

api-daemon/provisioner/examples/debootstrap_script.py

@@ -34,6 +34,29 @@
 # with that.
 
 import os
+from contextlib import contextmanager
+
+
+# Create a chroot context manager
+# This can be used later in the script to chroot to the destination directory
+# for instance to run commands within the target.
+@contextmanager
+def chroot_target(destination):
+    try:
+        real_root = os.open("/", os.O_RDONLY)
+        os.chroot(destination)
+        fake_root = os.open("/", os.O_RDONLY)
+        os.fchdir(fake_root)
+        yield
+    finally:
+        os.fchdir(real_root)
+        os.chroot(".")
+        os.fchdir(real_root)
+        os.close(fake_root)
+        os.close(real_root)
+        del fake_root
+        del real_root
+
 
 # Installation function - performs a debootstrap install of a Debian system
 # Note that the only arguments are keyword arguments.
@@ -193,13 +216,7 @@ GRUB_DISABLE_LINUX_UUID=false
         fh.write(data)
 
     # Chroot, do some in-root tasks, then exit the chroot
-    # EXITING THE CHROOT IS VERY IMPORTANT OR THE FOLLOWING STAGES OF THE PROVISIONER
-    # WILL FAIL IN UNEXPECTED WAYS! Keep this in mind when using chroot in your scripts.
-    real_root = os.open("/", os.O_RDONLY)
-    os.chroot(temporary_directory)
-    fake_root = os.open("/", os.O_RDONLY)
-    os.fchdir(fake_root)
-
+    with chroot_target(temporary_directory):
         # Install and update GRUB
         os.system(
             "grub-install --force /dev/rbd/{}/{}_{}".format(root_disk['pool'], vm_name, root_disk['disk_id'])
@@ -219,15 +236,6 @@ GRUB_DISABLE_LINUX_UUID=false
             "systemctl enable cloud-init.target"
         )
 
-    # Restore our original root/exit the chroot
-    # EXITING THE CHROOT IS VERY IMPORTANT OR THE FOLLOWING STAGES OF THE PROVISIONER
-    # WILL FAIL IN UNEXPECTED WAYS! Keep this in mind when using chroot in your scripts.
-    os.fchdir(real_root)
-    os.chroot(".")
-    os.fchdir(real_root)
-    os.close(fake_root)
-    os.close(real_root)
-
     # Unmount the bound devfs
     os.system(
         "umount {}/dev".format(
@@ -235,8 +243,4 @@ GRUB_DISABLE_LINUX_UUID=false
         )
     )
 
-    # Clean up file handles so paths can be unmounted
-    del fake_root
-    del real_root
-
     # Everything else is done via cloud-init user-data

api-daemon/provisioner/examples/dummy_script.py

@@ -29,7 +29,7 @@
 # This script will run under root privileges as the provisioner does. Be careful
 # with that.
 
-# Installation function - performs a debootstrap install of a Debian system
+# Installation function - performs no actions then returns
 # Note that the only arguments are keyword arguments.
 def install(**kwargs):
     # The provisioner has already mounted the disks on kwargs['temporary_directory'].

api-daemon/pvcapid/Daemon.py

@@ -25,7 +25,7 @@ import yaml
 from distutils.util import strtobool as dustrtobool
 
 # Daemon version
-version = '0.9.23'
+version = '0.9.25'
 
 # API version
 API_VERSION = 1.0

client-cli/pvc/pvc.py

@@ -251,7 +251,11 @@ def cluster_remove(name):
 # pvc cluster list
 ###############################################################################
 @click.command(name='list', short_help='List all available clusters.')
-def cluster_list():
+@click.option(
+    '-r', '--raw', 'raw', is_flag=True, default=False,
+    help='Display the raw list of cluster names only.'
+)
+def cluster_list(raw):
     """
     List all the available PVC clusters configured in this CLI instance.
     """
@@ -302,6 +306,7 @@ def cluster_list():
         if _api_key_length > api_key_length:
             api_key_length = _api_key_length
 
+    if not raw:
         # Display the data nicely
         click.echo("Available clusters:")
         click.echo()
@@ -341,6 +346,7 @@ def cluster_list():
             if not api_key:
                 api_key = 'N/A'
 
+        if not raw:
             click.echo(
                 '{bold}{name: <{name_length}} {description: <{description_length}} {address: <{address_length}} {port: <{port_length}} {scheme: <{scheme_length}} {api_key: <{api_key_length}}{end_bold}'.format(
                     bold='',
@@ -359,6 +365,8 @@ def cluster_list():
                     api_key_length=api_key_length
                 )
             )
+        else:
+            click.echo(cluster)
 
 
 # Validate that the cluster is set for a given command
@@ -1652,7 +1660,7 @@ def vm_dump(filename, domain):
 @cluster_req
 def vm_list(target_node, target_state, limit, raw):
     """
-    List all virtual machines; optionally only match names matching regex LIMIT.
+    List all virtual machines; optionally only match names or full UUIDs matching regex LIMIT.
 
     NOTE: Red-coloured network lists indicate one or more configured networks are missing/invalid.
     """

client-cli/setup.py

@@ -2,7 +2,7 @@ from setuptools import setup
 
 setup(
     name='pvc',
-    version='0.9.23',
+    version='0.9.25',
     packages=['pvc', 'pvc.cli_lib'],
     install_requires=[
         'Click',

daemon-common/common.py

@@ -343,8 +343,13 @@ def getInformationFromXML(zkhandler, uuid):
 
     parsed_xml = getDomainXML(zkhandler, uuid)
 
-    stats_data = loads(zkhandler.read(('domain.stats', uuid)))
-    if stats_data is None:
+    stats_data = zkhandler.read(('domain.stats', uuid))
+    if stats_data is not None:
+        try:
+            stats_data = loads(stats_data)
+        except Exception:
+            stats_data = {}
+    else:
         stats_data = {}
 
     domain_uuid, domain_name, domain_description, domain_memory, domain_vcpu, domain_vcputopo = getDomainMainDetails(parsed_xml)

daemon-common/vm.py

@@ -449,14 +449,6 @@ def remove_vm(zkhandler, domain):
     if current_vm_state != 'stop':
         change_state(zkhandler, dom_uuid, 'stop')
 
-    # Gracefully terminate the class instances
-    change_state(zkhandler, dom_uuid, 'delete')
-
-    # Delete the configurations
-    zkhandler.delete([
-        ('domain', dom_uuid)
-    ])
-
     # Wait for 1 second to allow state to flow to all nodes
     time.sleep(1)
 
@@ -465,11 +457,28 @@ def remove_vm(zkhandler, domain):
         # vmpool/vmname_volume
         try:
             disk_pool, disk_name = disk.split('/')
-            retcode, message = ceph.remove_volume(zkhandler, disk_pool, disk_name)
         except ValueError:
             continue
 
-    return True, 'Removed VM "{}" and disks from the cluster.'.format(domain)
+        retcode, message = ceph.remove_volume(zkhandler, disk_pool, disk_name)
+        if not retcode:
+            if re.match('^ERROR: No volume with name', message):
+                continue
+            else:
+                return False, message
+
+    # Gracefully terminate the class instances
+    change_state(zkhandler, dom_uuid, 'delete')
+
+    # Wait for 1/2 second to allow state to flow to all nodes
+    time.sleep(0.5)
+
+    # Delete the VM configuration from Zookeeper
+    zkhandler.delete([
+        ('domain', dom_uuid)
+    ])
+
+    return True, 'Removed VM "{}" and its disks from the cluster.'.format(domain)
 
 
 def start_vm(zkhandler, domain):
@@ -789,7 +798,10 @@ def get_console_log(zkhandler, domain, lines=1000):
         return False, 'ERROR: Could not find VM "{}" in the cluster!'.format(domain)
 
     # Get the data from ZK
-    console_log = zkhandler.read(('domain.log.console', dom_uuid))
+    console_log = zkhandler.read(('domain.console.log', dom_uuid))
+
+    if console_log is None:
+        return True, ''
 
     # Shrink the log buffer to length lines
     shrunk_log = console_log.split('\n')[-lines:]
@@ -897,6 +909,9 @@ def get_list(zkhandler, node, state, limit, is_fuzzy=True):
         for vm_uuid in vm_execute_list:
             futures.append(executor.submit(common.getInformationFromXML, zkhandler, vm_uuid))
         for future in futures:
+            try:
                 vm_data_list.append(future.result())
+            except Exception:
+                pass
 
     return True, vm_data_list

daemon-common/zkhandler.py

@@ -124,37 +124,29 @@ class ZKHandler(object):
     # State/connection management
     #
     def listener(self, state):
+        """
+        Listen for KazooState changes and log accordingly.
+
+        This function does not do anything except for log the state, and Kazoo handles the rest.
+        """
         if state == KazooState.CONNECTED:
-            self.log('Connection to Zookeeper started', state='o')
+            self.log('Connection to Zookeeper resumed', state='o')
         else:
-            self.log('Connection to Zookeeper lost', state='w')
-
-            while True:
-                time.sleep(0.5)
-
-                _zk_conn = KazooClient(hosts=self.coordinators)
-                try:
-                    _zk_conn.start()
-                except Exception:
-                    del _zk_conn
-                    continue
-
-                self.zk_conn = _zk_conn
-                self.zk_conn.add_listener(self.listener)
-                break
+            self.log('Connection to Zookeeper lost with state {}'.format(state), state='w')
 
     def connect(self, persistent=False):
         """
-        Start the zk_conn object and connect to the cluster, then load the current schema version
+        Start the zk_conn object and connect to the cluster
         """
         try:
             self.zk_conn.start()
             if persistent:
+                self.log('Connection to Zookeeper started', state='o')
                 self.zk_conn.add_listener(self.listener)
         except Exception as e:
             raise ZKConnectionException(self, e)
 
-    def disconnect(self):
+    def disconnect(self, persistent=False):
         """
         Stop and close the zk_conn object and disconnect from the cluster
 
@@ -162,11 +154,27 @@ class ZKHandler(object):
         """
         self.zk_conn.stop()
         self.zk_conn.close()
+        if persistent:
+            self.log('Connection to Zookeeper terminated', state='o')
 
     #
     # Schema helper actions
     #
     def get_schema_path(self, key):
+        """
+        Get the Zookeeper path for {key} from the current schema based on its format.
+
+        If {key} is a tuple of length 2, it's treated as a path plus an item instance of that path (e.g. a node, a VM, etc.).
+
+        If {key} is a tuple of length 4, it is treated as a path plus an item instance, as well as another item instance of the subpath.
+
+        If {key} is just a string, it's treated as a lone path (mostly used for the 'base' schema group.
+
+        Otherwise, returns None since this is not a valid key.
+
+        This function also handles the special case where a string that looks like an existing path (i.e. starts with '/') is passed;
+        in that case it will silently return the same path back. This was mostly a migration functionality and is deprecated.
+        """
         if isinstance(key, tuple):
             # This is a key tuple with both an ipath and an item
             if len(key) == 2:

debian/changelog

@@ -1,3 +1,23 @@
+pvc (0.9.25-0) unstable; urgency=high
+
+  * [Node Daemon] Returns to Rados library calls for Ceph due to performance problems
+  * [Node Daemon] Adds a date output to keepalive messages
+  * [Daemons] Configures ZK connection logging only for persistent connections
+  * [API Provisioner] Add context manager-based chroot to Debootstrap example script
+  * [Node Daemon] Fixes a bug where shutdown daemon state was overwritten
+
+ -- Joshua M. Boniface <joshua@boniface.me>  Sun, 11 Jul 2021 23:19:09 -0400
+
+pvc (0.9.24-0) unstable; urgency=high
+
+  * [Node Daemon] Removes Rados module polling of Ceph cluster and returns to command-based polling for timeout purposes, and removes some flaky return statements
+  * [Node Daemon] Removes flaky Zookeeper connection renewals that caused problems
+  * [CLI Client] Allow raw lists of clusters from `pvc cluster list`
+  * [API Daemon] Fixes several issues when getting VM data without stats
+  * [API Daemon] Fixes issues with removing VMs while disks are still in use (failed provisioning, etc.)
+
+ -- Joshua M. Boniface <joshua@boniface.me>  Fri, 09 Jul 2021 15:58:36 -0400
+
 pvc (0.9.23-0) unstable; urgency=high
 
   * [Daemons] Fixes a critical overwriting bug in zkhandler when schema paths are not yet valid

docs/index.md

@@ -42,6 +42,22 @@ To get started with PVC, please see the [About](https://parallelvirtualcluster.r
 
 ## Changelog
 
+#### v0.9.25
+
+  * [Node Daemon] Returns to Rados library calls for Ceph due to performance problems
+  * [Node Daemon] Adds a date output to keepalive messages
+  * [Daemons] Configures ZK connection logging only for persistent connections
+  * [API Provisioner] Add context manager-based chroot to Debootstrap example script
+  * [Node Daemon] Fixes a bug where shutdown daemon state was overwritten
+
+#### v0.9.24
+
+  * [Node Daemon] Removes Rados module polling of Ceph cluster and returns to command-based polling for timeout purposes, and removes some flaky return statements
+  * [Node Daemon] Removes flaky Zookeeper connection renewals that caused problems
+  * [CLI Client] Allow raw lists of clusters from `pvc cluster list`
+  * [API Daemon] Fixes several issues when getting VM data without stats
+  * [API Daemon] Fixes issues with removing VMs while disks are still in use (failed provisioning, etc.)
+
 #### v0.9.23
 
   * [Daemons] Fixes a critical overwriting bug in zkhandler when schema paths are not yet valid

node-daemon/pvcnoded/Daemon.py

@@ -32,6 +32,7 @@ import yaml
 import json
 
 from socket import gethostname
+from datetime import datetime
 from threading import Thread
 from ipaddress import ip_address, ip_network
 from apscheduler.schedulers.background import BackgroundScheduler
@@ -55,7 +56,7 @@ import pvcnoded.CephInstance as CephInstance
 import pvcnoded.MetadataAPIInstance as MetadataAPIInstance
 
 # Version string for startup output
-version = '0.9.23'
+version = '0.9.25'
 
 ###############################################################################
 # PVCD - node daemon startup program
@@ -658,7 +659,7 @@ def update_schema(new_schema_version, stat, event=''):
     # Restart ourselves with the new schema
     logger.out('Reloading node daemon', state='s')
     try:
-        zkhandler.disconnect()
+        zkhandler.disconnect(persistent=True)
         del zkhandler
     except Exception:
         pass
@@ -751,7 +752,7 @@ def cleanup():
 
     # Close the Zookeeper connection
     try:
-        zkhandler.disconnect()
+        zkhandler.disconnect(persistent=True)
         del zkhandler
     except Exception:
         pass
@@ -1334,11 +1335,13 @@ def collect_ceph_stats(queue):
         ceph_health = health_status['status']
     except Exception as e:
         logger.out('Failed to obtain Ceph health data: {}'.format(e), state='e')
-        return
+        ceph_health = 'HEALTH_UNKN'
 
-    if ceph_health == 'HEALTH_OK':
+    if ceph_health in ['HEALTH_OK']:
         ceph_health_colour = fmt_green
-    elif ceph_health == 'HEALTH_WARN':
+    elif ceph_health in ['HEALTH_UNKN']:
+        ceph_health_colour = fmt_cyan
+    elif ceph_health in ['HEALTH_WARN']:
         ceph_health_colour = fmt_yellow
     else:
         ceph_health_colour = fmt_red
@@ -1356,7 +1359,6 @@ def collect_ceph_stats(queue):
             ])
         except Exception as e:
             logger.out('Failed to set Ceph status data: {}'.format(e), state='e')
-            return
 
         if debug:
             logger.out("Set ceph rados df information in zookeeper (primary only)", state='d', prefix='ceph-thread')
@@ -1370,15 +1372,15 @@ def collect_ceph_stats(queue):
             ])
         except Exception as e:
             logger.out('Failed to set Ceph utilization data: {}'.format(e), state='e')
-            return
 
         if debug:
             logger.out("Set pool information in zookeeper (primary only)", state='d', prefix='ceph-thread')
 
         # Get pool info
-        retcode, stdout, stderr = common.run_os_command('ceph df --format json', timeout=1)
+        command = {"prefix": "df", "format": "json"}
+        ceph_df_output = ceph_conn.mon_command(json.dumps(command), b'', timeout=1)[1].decode('ascii')
         try:
-            ceph_pool_df_raw = json.loads(stdout)['pools']
+            ceph_pool_df_raw = json.loads(ceph_df_output)['pools']
         except Exception as e:
             logger.out('Failed to obtain Pool data (ceph df): {}'.format(e), state='w')
             ceph_pool_df_raw = []
@@ -1449,9 +1451,9 @@ def collect_ceph_stats(queue):
         osd_dump = dict()
 
         command = {"prefix": "osd dump", "format": "json"}
+        osd_dump_output = ceph_conn.mon_command(json.dumps(command), b'', timeout=1)[1].decode('ascii')
         try:
-            retcode, stdout, stderr = common.run_os_command('ceph osd dump --format json --connect-timeout 2', timeout=2)
-            osd_dump_raw = json.loads(stdout)['osds']
+            osd_dump_raw = json.loads(osd_dump_output)['osds']
         except Exception as e:
             logger.out('Failed to obtain OSD data: {}'.format(e), state='w')
             osd_dump_raw = []
@@ -1608,7 +1610,6 @@ def collect_vm_stats(queue):
     lv_conn = libvirt.open(libvirt_name)
     if lv_conn is None:
         logger.out('Failed to open connection to "{}"'.format(libvirt_name), state='e')
-        return
 
     memalloc = 0
     memprov = 0
@@ -1778,8 +1779,9 @@ def node_keepalive():
     # Get past state and update if needed
     if debug:
         logger.out("Get past state and update if needed", state='d', prefix='main-thread')
+
     past_state = zkhandler.read(('node.state.daemon', this_node.name))
-    if past_state != 'run':
+    if past_state != 'run' and past_state != 'shutdown':
         this_node.daemon_state = 'run'
         zkhandler.write([
             (('node.state.daemon', this_node.name), 'run')
@@ -1868,7 +1870,6 @@ def node_keepalive():
         ])
     except Exception:
         logger.out('Failed to set keepalive data', state='e')
-        return
 
     # Display node information to the terminal
     if config['log_keepalives']:
@@ -1879,9 +1880,10 @@ def node_keepalive():
         else:
             cst_colour = fmt_cyan
         logger.out(
-            '{}{} keepalive{} [{}{}{}]'.format(
+            '{}{} keepalive @ {}{} [{}{}{}]'.format(
                 fmt_purple,
                 myhostname,
+                datetime.now(),
                 fmt_end,
                 fmt_bold + cst_colour,
                 this_node.router_state,