Merge branch 'main' into metrics

config.yaml

@@ -21,6 +21,12 @@ options:
     description: |
       Channel to install the DCGM snap if the hardware has NVIDIA GPU. By default, it will install
       from latest/stable
+  smartctl-exporter-snap-channel:
+    type: string
+    default: "latest/stable"
+    description: |
+      Channel to install the Smartctl exporter snap if the hardware has smart disk. By default, it will install
+      from latest/stable.
   exporter-log-level:
     type: string
     default: "INFO"

src/charm.py

@@ -12,7 +12,7 @@
 from ops.framework import EventBase, StoredState
 from ops.model import ActiveStatus, BlockedStatus, MaintenanceStatus
 
-from hw_tools import HWTool, HWToolHelper, detect_available_tools
+from hw_tools import HWTool, HWToolHelper, detect_available_tools, remove_legacy_smartctl_exporter
 from service import BaseExporter, DCGMExporter, ExporterError, HardwareExporter, SmartCtlExporter
 
 logger = logging.getLogger(__name__)
@@ -81,7 +81,7 @@ def exporters(self) -> List[BaseExporter]:
             )
 
         if stored_tools & SmartCtlExporter.hw_tools():
-            exporters.append(SmartCtlExporter(self.charm_dir, self.model.config))
+            exporters.append(SmartCtlExporter(self.model.config))
 
         if stored_tools & DCGMExporter.hw_tools():
             exporters.append(DCGMExporter(self.charm_dir, self.model.config))
@@ -97,6 +97,8 @@ def get_stored_tools(self) -> Set[HWTool]:
         if not self._stored.stored_tools:  # type: ignore[truthy-function]
             available_tools = detect_available_tools()  # type: ignore[unreachable]
             self._stored.stored_tools = {tool.value for tool in available_tools}
+        if "smartctl" in self._stored.stored_tools:  # type: ignore[operator]
+            self._stored.stored_tools.remove("smartctl")  # type: ignore[attr-defined]
         return {HWTool(value) for value in self._stored.stored_tools}  # type: ignore[attr-defined]
 
     def _on_redetect_hardware(self, event: ops.ActionEvent) -> None:
@@ -130,6 +132,8 @@ def _on_install_or_upgrade(self, event: EventBase) -> None:
         """Install or upgrade charm."""
         self.model.unit.status = MaintenanceStatus("Installing resources...")
 
+        remove_legacy_smartctl_exporter()
+
         stored_tools = self.get_stored_tools()
 
         msg: str

src/config.py

@@ -36,20 +36,6 @@ class HardwareExporterSettings(ExporterSettings):  # pylint: disable = too-few-p
 HARDWARE_EXPORTER_SETTINGS = HardwareExporterSettings()
 
 
-class SmartCtlExporterSettings(ExporterSettings):  # pylint: disable = too-few-public-methods
-    """Constant settings for SmartCtl Exporter."""
-
-    name: str = "smartctl-exporter"
-    config_path: Path = Path(f"/etc/{name}-config.yaml")
-    service_path: Path = Path(f"/etc/systemd/system/{name}.service")
-    config_template: str = f"{name}-config.yaml.j2"
-    service_template: str = f"{name}.service.j2"
-    crash_msg: str = "SmartCtl exporter crashed unexpectedly, please refer to systemd logs..."
-
-
-SMARTCTL_EXPORTER_SETTINGS = SmartCtlExporterSettings()
-
-
 class SystemVendor(str, Enum):
     """Different hardware system vendor."""
 
@@ -77,8 +63,7 @@ class HWTool(str, Enum):
     IPMI_SEL = "ipmi_sel"
     IPMI_SENSOR = "ipmi_sensor"
     REDFISH = "redfish"
-    SMARTCTL = "smartctl"
-    SMARTCTL_EXPORTER = "smartctl_exporter"
+    SMARTCTL_EXPORTER = "smartctl-exporter"
     DCGM = "dcgm"
 
 

src/hw_tools.py

@@ -3,21 +3,19 @@
 Define strategy for install, remove and verifier for different hardware.
 """
 
-import io
 import logging
 import os
 import shutil
 import stat
 import subprocess
-import tarfile
 from abc import ABCMeta, abstractmethod
-from http import HTTPStatus
 from pathlib import Path
 from typing import Dict, List, Set, Tuple
 
 import requests
 import urllib3
 from charms.operator_libs_linux.v0 import apt
+from charms.operator_libs_linux.v1 import systemd
 from charms.operator_libs_linux.v2 import snap
 from ops.model import ModelError, Resources
 
@@ -63,14 +61,6 @@ def __init__(self, tool: HWTool, path: Path):
         self.message = f"Tool: {tool} path: {path} size is zero"
 
 
-class ResourceInstallationError(Exception):
-    """Exception raised when a hardware tool installation fails."""
-
-    def __init__(self, tool: HWTool):
-        """Init."""
-        super().__init__(f"Installation failed for tool: {tool}")
-
-
 def copy_to_snap_common_bin(source: Path, filename: str) -> None:
     """Copy file to $SNAP_COMMON/bin folder."""
     Path(f"{SNAP_COMMON}/bin").mkdir(parents=False, exist_ok=True)
@@ -239,6 +229,16 @@ def __init__(self, channel: str) -> None:
         self.channel = channel
 
 
+class SmartCtlExporterStrategy(SnapStrategy):
+    """SmartCtl strategy class."""
+
+    _name = HWTool.SMARTCTL_EXPORTER
+
+    def __init__(self, channel: str) -> None:
+        """Init."""
+        self.channel = channel
+
+
 class StorCLIStrategy(TPRStrategyABC):
     """Strategy to install storcli."""
 
@@ -444,75 +444,6 @@ def check(self) -> bool:
         return True
 
 
-class SmartCtlStrategy(APTStrategyABC):
-    """Strategy for installing ipmi."""
-
-    pkg = "smartmontools"
-    _name = HWTool.SMARTCTL
-
-    def install(self) -> None:
-        apt_helpers.add_pkg_with_candidate_version(self.pkg)
-
-    def remove(self) -> None:
-        # Skip removing because this may cause dependency error
-        # for other services on the same machine.
-        logger.info("%s skip removing %s", self._name, self.pkg)
-
-    def check(self) -> bool:
-        """Check package status."""
-        return check_deb_pkg_installed(self.pkg)
-
-
-class SmartCtlExporterStrategy(StrategyABC):  # pylint: disable=R0903
-    """Install smartctl exporter binary."""
-
-    _name = HWTool.SMARTCTL_EXPORTER
-
-    _resource_dir = Path("/opt/SmartCtlExporter/")
-    _release = (
-        "https://github.com/prometheus-community/"
-        "smartctl_exporter/releases/download/v0.12.0/smartctl_exporter-0.12.0.linux-amd64.tar.gz"
-    )
-    _exporter_name = "smartctl_exporter"
-    _exporter_path = Path(_resource_dir / "smartctl_exporter")
-
-    def install(self) -> None:
-        """Install exporter binary from internet."""
-        logger.debug("Installing SmartCtlExporter")
-        self._resource_dir.mkdir(parents=True, exist_ok=True)
-
-        resp = requests.get(self._release, timeout=60)
-        if resp.status_code != HTTPStatus.OK:
-            logger.error("Failed to download smartctl exporter binary.")
-            raise ResourceInstallationError(self._name)
-
-        success = False
-        fileobj = io.BytesIO(resp.content)
-        with tarfile.open(fileobj=fileobj, mode="r:gz") as tar:
-            for member in tar.getmembers():
-                if member.name.endswith(self._exporter_name):
-                    with open(self._exporter_path, "wb") as outfile:
-                        member_file = tar.extractfile(member)
-                        if member_file:
-                            outfile.write(member_file.read())
-                            success = True
-                    if success:
-                        make_executable(self._exporter_path)
-        if not success:
-            logger.error("Failed to install SmartCtlExporter binary.")
-            raise ResourceInstallationError(self._name)
-
-    def remove(self) -> None:
-        """Remove downloaded exporter binary."""
-        logger.debug("Remove SmartCtlExporter")
-        shutil.rmtree(self._resource_dir)
-
-    def check(self) -> bool:
-        """Check package status."""
-        logger.debug("Check SmartCtlExporter resources")
-        return self._exporter_path.is_file()
-
-
 def _raid_hw_verifier_hwinfo() -> Set[HWTool]:
     """Verify if a supported RAID card exists on the machine using the hwinfo command."""
     hwinfo_output = hwinfo("storage")
@@ -650,7 +581,7 @@ def bmc_hw_verifier() -> Set[HWTool]:
 
 def disk_hw_verifier() -> Set[HWTool]:
     """Verify if the disk exists on the machine."""
-    return {HWTool.SMARTCTL} if lshw(class_filter="disk") else set()
+    return {HWTool.SMARTCTL_EXPORTER} if lshw(class_filter="disk") else set()
 
 
 def nvidia_gpu_verifier() -> Set[HWTool]:
@@ -664,6 +595,25 @@ def detect_available_tools() -> Set[HWTool]:
     return raid_hw_verifier() | bmc_hw_verifier() | disk_hw_verifier() | nvidia_gpu_verifier()
 
 
+def remove_legacy_smartctl_exporter() -> None:
+    """Remove any legacy tool from older revision.
+
+    Workaround for migrating legacy smartctl exporter to snap package.
+    """
+    name = "smartctl-exporter"
+    smartctl_exporter = Path("opt/SmartCtlExporter/")
+    smartctl_exporter_config_path = Path(f"/etc/{name}-config.yaml")
+    smartctl_exporter_service_path = Path(f"/etc/systemd/system/{name}.service")
+    if smartctl_exporter_service_path.exists():
+        systemd.service_stop(name)
+        systemd.service_disable(name)
+        smartctl_exporter_service_path.unlink()
+    if smartctl_exporter_config_path.exists():
+        smartctl_exporter_config_path.unlink()
+    if smartctl_exporter.exists():
+        shutil.rmtree("/opt/SmartCtlExporter/")
+
+
 class HWToolHelper:
     """Helper to install vendor's or hardware related tools."""
 
@@ -680,7 +630,6 @@ def strategies(self) -> List[StrategyABC]:
             IPMIDCMIStrategy(),
             IPMISENSORStrategy(),
             RedFishStrategy(),
-            SmartCtlStrategy(),
         ]
 
     def fetch_tools(  # pylint: disable=W0102