Improve fence handling conditions

Use the intermediate output text when judging the fence status, rather
than the retcode of the stop as this should be more reliable.

.version

@@ -1 +1 @@
-0.9.96
+0.9.97

CHANGELOG.md

@@ -1,5 +1,12 @@
 ## PVC Changelog
 
+###### [v0.9.97](https://github.com/parallelvirtualcluster/pvc/releases/tag/v0.9.97)
+
+  * [Client CLI] Ensures --lines is always an integer value
+  * [Node Daemon] Fixes a bug if d_network changes during iteration
+  * [Node Daemon] Moves to using allocated instead of free memory for node reporting
+  * [API Daemon] Fixes a bug if lingering RBD snapshots exist when removing a volume (#180)
+
 ###### [v0.9.96](https://github.com/parallelvirtualcluster/pvc/releases/tag/v0.9.96)
 
   * [API Daemon] Fixes a bug when reporting node stats

api-daemon/pvcapid/Daemon.py

@@ -27,7 +27,7 @@ from distutils.util import strtobool as dustrtobool
 import daemon_lib.config as cfg
 
 # Daemon version
-version = "0.9.96"
+version = "0.9.97"
 
 # API version
 API_VERSION = 1.0

client-cli/pvc/cli/cli.py

@@ -671,9 +671,9 @@ def cli_cluster_maintenance_off():
 @format_opt(
     {
         "pretty": cli_cluster_task_format_pretty,
-        "raw": lambda d: "\n".join([t["id"] for t in d])
-        if isinstance(d, list)
-        else d["state"],
+        "raw": lambda d: (
+            "\n".join([t["id"] for t in d]) if isinstance(d, list) else d["state"]
+        ),
         "json": lambda d: jdumps(d),
         "json-pretty": lambda d: jdumps(d, indent=2),
     }
@@ -892,6 +892,7 @@ def cli_node_ready(
     "--lines",
     "lines",
     default=None,
+    type=int,
     show_default=False,
     help="Display this many log lines from the end of the log buffer.  [default: 1000; with follow: 10]",
 )
@@ -2516,6 +2517,7 @@ def cli_vm_volume_remove(domain, volume, live_flag, restart_flag):
     "--lines",
     "lines",
     default=None,
+    type=int,
     show_default=False,
     help="Display this many log lines from the end of the log buffer.  [default: 1000; with follow: 10]",
 )

client-cli/pvc/cli/formatters.py

@@ -580,9 +580,11 @@ def cli_cluster_fault_list_format_long(CLI_CONFIG, fault_data):
                 fault_id=fault["id"],
                 fault_status=fault["status"].title(),
                 fault_health_delta=f"-{fault['health_delta']}%",
-                fault_acknowledged_at=fault["acknowledged_at"]
-                if fault["acknowledged_at"] != ""
-                else "N/A",
+                fault_acknowledged_at=(
+                    fault["acknowledged_at"]
+                    if fault["acknowledged_at"] != ""
+                    else "N/A"
+                ),
                 fault_last_reported=fault["last_reported"],
                 fault_first_reported=fault["first_reported"],
             )

client-cli/pvc/lib/common.py

@@ -108,9 +108,10 @@ class UploadProgressBar(object):
 
 
 class ErrorResponse(requests.Response):
-    def __init__(self, json_data, status_code):
+    def __init__(self, json_data, status_code, headers):
         self.json_data = json_data
         self.status_code = status_code
+        self.headers = headers
 
     def json(self):
         return self.json_data
@@ -206,7 +207,7 @@ def call_api(
     except Exception as e:
         message = "Failed to connect to the API: {}".format(e)
         code = response.status_code if response else 504
-        response = ErrorResponse({"message": message}, code)
+        response = ErrorResponse({"message": message}, code, None)
 
     # Display debug output
     if config["debug"]:

client-cli/pvc/lib/vm.py

@@ -1765,9 +1765,9 @@ def format_info(config, domain_information, long_output):
                     tags_name=tag["name"],
                     tags_type=tag["type"],
                     tags_protected=str(tag["protected"]),
-                    tags_protected_colour=ansiprint.green()
-                    if tag["protected"]
-                    else ansiprint.blue(),
+                    tags_protected_colour=(
+                        ansiprint.green() if tag["protected"] else ansiprint.blue()
+                    ),
                     end=ansiprint.end(),
                 )
             )

client-cli/setup.py

@@ -2,7 +2,7 @@ from setuptools import setup
 
 setup(
     name="pvc",
-    version="0.9.96",
+    version="0.9.97",
     packages=["pvc.cli", "pvc.lib"],
     install_requires=[
         "Click",

daemon-common/ceph.py

@@ -320,7 +320,11 @@ def get_list_osd(zkhandler, limit=None, is_fuzzy=True):
 #
 def getPoolInformation(zkhandler, pool):
     # Parse the stats data
-    (pool_stats_raw, tier, pgs,) = zkhandler.read_many(
+    (
+        pool_stats_raw,
+        tier,
+        pgs,
+    ) = zkhandler.read_many(
         [
             ("pool.stats", pool),
             ("pool.tier", pool),
@@ -824,10 +828,22 @@ def remove_volume(zkhandler, pool, name):
             name, pool
         )
 
-    # 1. Remove volume snapshots
+    # 1a. Remove PVC-managed volume snapshots
     for snapshot in zkhandler.children(("snapshot", f"{pool}/{name}")):
         remove_snapshot(zkhandler, pool, name, snapshot)
 
+    # 1b. Purge any remaining volume snapshots
+    retcode, stdout, stderr = common.run_os_command(
+        "rbd snap purge {}/{}".format(pool, name)
+    )
+    if retcode:
+        return (
+            False,
+            'ERROR: Failed to purge snapshots from RBD volume "{}" in pool "{}": {}'.format(
+                name, pool, stderr
+            ),
+        )
+
     # 2. Remove the volume
     retcode, stdout, stderr = common.run_os_command("rbd rm {}/{}".format(pool, name))
     if retcode:

daemon-common/config.py

@@ -244,9 +244,9 @@ def get_parsed_configuration(config_file):
                     ]
                 ][0]
 
-            config_cluster_networks_specific[
-                f"{network_type}_dev_ip"
-            ] = f"{list(network.hosts())[address_id]}/{network.prefixlen}"
+            config_cluster_networks_specific[f"{network_type}_dev_ip"] = (
+                f"{list(network.hosts())[address_id]}/{network.prefixlen}"
+            )
 
             config = {**config, **config_cluster_networks_specific}
 

debian/changelog

@@ -1,3 +1,12 @@
+pvc (0.9.97-0) unstable; urgency=high
+
+  * [Client CLI] Ensures --lines is always an integer value
+  * [Node Daemon] Fixes a bug if d_network changes during iteration
+  * [Node Daemon] Moves to using allocated instead of free memory for node reporting
+  * [API Daemon] Fixes a bug if lingering RBD snapshots exist when removing a volume (#180)
+
+ -- Joshua M. Boniface <joshua@boniface.me>  Fri, 19 Apr 2024 10:32:16 -0400
+
 pvc (0.9.96-0) unstable; urgency=high
 
   * [API Daemon] Fixes a bug when reporting node stats

health-daemon/pvchealthd/Daemon.py

@@ -33,7 +33,7 @@ import os
 import signal
 
 # Daemon version
-version = "0.9.96"
+version = "0.9.97"
 
 
 ##########################################################

node-daemon/pvcnoded/Daemon.py

@@ -49,7 +49,7 @@ import re
 import json
 
 # Daemon version
-version = "0.9.96"
+version = "0.9.97"
 
 
 ##########################################################

node-daemon/pvcnoded/objects/NetstatsInstance.py

@@ -231,7 +231,7 @@ class NetstatsInstance(object):
         # Get a list of all active interfaces
         net_root_path = "/sys/class/net"
         all_ifaces = list()
-        for (_, dirnames, _) in walk(net_root_path):
+        for _, dirnames, _ in walk(net_root_path):
             all_ifaces.extend(dirnames)
         all_ifaces.sort()
 

node-daemon/pvcnoded/objects/NodeInstance.py

@@ -521,7 +521,7 @@ class NodeInstance(object):
         self.logger.out("Acquired write lock for synchronization phase F", state="o")
         time.sleep(0.2)  # Time fir reader to acquire the lock
         # 4. Add gateway IPs
-        for network in self.d_network:
+        for network in self.d_network.copy():
             self.d_network[network].createGateways()
         self.logger.out("Releasing write lock for synchronization phase F", state="i")
         self.zkhandler.write([("base.config.primary_node.sync_lock", "")])

node-daemon/pvcnoded/util/fencing.py

@@ -253,12 +253,16 @@ def reboot_via_ipmi(node_name, ipmi_hostname, ipmi_user, ipmi_password, logger):
         state="i",
         prefix=f"fencing {node_name}",
     )
-    ipmi_status_retcode, ipmi_status_stdout, ipmi_status_stderr = common.run_os_command(
+    (
+        ipmi_intermediate_status_retcode,
+        ipmi_intermediate_status_stdout,
+        ipmi_intermediate_status_stderr,
+    ) = common.run_os_command(
         f"/usr/bin/ipmitool -I lanplus -H {ipmi_hostname} -U {ipmi_user} -P {ipmi_password} chassis power status"
     )
-    if ipmi_status_retcode == 0:
+    if ipmi_intermediate_status_retcode == 0:
         logger.out(
-            f"Current chassis power state is: {ipmi_status_stdout.strip()}",
+            f"Current chassis power state is: {ipmi_intermediate_status_stdout.strip()}",
             state="i",
             prefix=f"fencing {node_name}",
         )
@@ -299,12 +303,14 @@ def reboot_via_ipmi(node_name, ipmi_hostname, ipmi_user, ipmi_password, logger):
         state="i",
         prefix=f"fencing {node_name}",
     )
-    ipmi_status_retcode, ipmi_status_stdout, ipmi_status_stderr = common.run_os_command(
-        f"/usr/bin/ipmitool -I lanplus -H {ipmi_hostname} -U {ipmi_user} -P {ipmi_password} chassis power status"
+    ipmi_final_status_retcode, ipmi_final_status_stdout, ipmi_final_status_stderr = (
+        common.run_os_command(
+            f"/usr/bin/ipmitool -I lanplus -H {ipmi_hostname} -U {ipmi_user} -P {ipmi_password} chassis power status"
+        )
     )
 
-    if ipmi_stop_retcode == 0:
-        if ipmi_status_stdout.strip() == "Chassis Power is on":
+    if ipmi_intermediate_status_stdout.strip() == "Chassis power is off":
+        if ipmi_final_status_stdout.strip() == "Chassis Power is on":
             # We successfully rebooted the node and it is powered on; this is a succeessful fence
             logger.out(
                 "Successfully rebooted dead node; proceeding with fence recovery action",
@@ -312,7 +318,7 @@ def reboot_via_ipmi(node_name, ipmi_hostname, ipmi_user, ipmi_password, logger):
                 prefix=f"fencing {node_name}",
             )
             return True
-        elif ipmi_status_stdout.strip() == "Chassis Power is off":
+        elif ipmi_final_status_stdout.strip() == "Chassis Power is off":
             # We successfully rebooted the node but it is powered off; this might be expected or not, but the node is confirmed off so we can call it a successful fence
             logger.out(
                 "Chassis power is in confirmed off state after successfuly IPMI reboot; proceeding with fence recovery action",
@@ -323,13 +329,13 @@ def reboot_via_ipmi(node_name, ipmi_hostname, ipmi_user, ipmi_password, logger):
         else:
             # We successfully rebooted the node but it is in some unknown power state; since this might indicate a silent failure, we must call it a failed fence
             logger.out(
-                f"Chassis power is in an unknown state ({ipmi_status_stdout.strip()}) after successful IPMI reboot; NOT proceeding fence recovery action",
+                f"Chassis power is in an unknown state ({ipmi_final_status_stdout.strip()}) after successful IPMI reboot; NOT proceeding fence recovery action",
                 state="e",
                 prefix=f"fencing {node_name}",
             )
             return False
     else:
-        if ipmi_status_stdout.strip() == "Chassis Power is off":
+        if ipmi_final_status_stdout.strip() == "Chassis Power is off":
             # We failed to reboot the node but it is powered off; it has probably suffered a serious hardware failure, but the node is confirmed off so we can call it a successful fence
             logger.out(
                 "Chassis power is in confirmed off state after failed IPMI reboot; proceeding with fence recovery action",

node-daemon/pvcnoded/util/keepalive.py

@@ -743,7 +743,7 @@ def node_keepalive(logger, config, zkhandler, this_node, netstats):
     # Get node performance statistics
     this_node.memtotal = int(psutil.virtual_memory().total / 1024 / 1024)
     this_node.memused = int(psutil.virtual_memory().used / 1024 / 1024)
-    this_node.memfree = int(psutil.virtual_memory().free / 1024 / 1024)
+    this_node.memfree = int(psutil.virtual_memory().available / 1024 / 1024)
     this_node.cpuload = round(os.getloadavg()[0], 2)
 
     # Get node network statistics via netstats instance

worker-daemon/pvcworkerd/Daemon.py

@@ -44,7 +44,7 @@ from daemon_lib.vmbuilder import (
 )
 
 # Daemon version
-version = "0.9.96"
+version = "0.9.97"
 
 
 config = cfg.get_configuration()