Merge branch 'main' of code.lotusguild.org:LotusGuild/analyzeOSDs

ceph_osd_analyzer.py

@@ -82,54 +82,66 @@ def get_osd_host_mapping(osd_tree):
     return osd_to_host
 
 def get_device_path_for_osd(osd_id, hostname):
-    """Get the device path for an OSD on a specific host"""
+    """Get the physical device path for an OSD on a host (resolve dm devices)."""
-    # Method 1: Try ceph metadata
     metadata = get_osd_metadata(osd_id)
     if metadata:
-        devices = metadata.get('devices', '')
+        # Try 'bluestore_bdev_devices' first
-        if devices:
+        phys_dev = metadata.get('bluestore_bdev_devices')
-            device = devices.split(',')[0] if ',' in devices else devices
+        if phys_dev:
-            if device and not device.startswith('/dev/'):
+            device = f"/dev/{phys_dev.strip()}"
-                device = f"/dev/{device}"
+            if DEBUG:
-            if device and device != '/dev/':
+                print(f"{Colors.GREEN}DEBUG: Found physical device from metadata: {device}{Colors.END}")
-                if DEBUG:
+            return device
-                    print(f"{Colors.GREEN}DEBUG: Found device from metadata: {device}{Colors.END}")
-                return device
     
-    # Method 2: Query symlink on remote host
+    # Fallback: follow the symlink
-    result = run_command(f"readlink -f /var/lib/ceph/osd/ceph-{osd_id}/block 2>/dev/null", host=hostname)
+    result = run_command(f"readlink -f /var/lib/ceph/osd/ceph-{osd_id}/block", host=hostname)
     if result and result.startswith('/dev/'):
-        if DEBUG:
+        # Check if it is a dm device, try to find underlying
-            print(f"{Colors.GREEN}DEBUG: Found device from symlink: {result}{Colors.END}")
+        if '/dev/dm-' in result:
-        return result
+            base = run_command(f"lsblk -no pkname {result}", host=hostname)
+            if base:
+                device = f"/dev/{base.strip()}"
+                if DEBUG:
+                    print(f"{Colors.GREEN}DEBUG: Resolved dm device {result} -> {device}{Colors.END}")
+                return device
+        else:
+            if DEBUG:
+                print(f"{Colors.GREEN}DEBUG: Using device symlink {result}{Colors.END}")
+            return result
     
-    # Method 3: Try lsblk
+    # Last fallback: lsblk from block path
-    result = run_command(f"lsblk -no pkname /var/lib/ceph/osd/ceph-{osd_id}/block 2>/dev/null", host=hostname)
+    result = run_command(f"lsblk -no pkname /var/lib/ceph/osd/ceph-{osd_id}/block", host=hostname)
     if result:
         device = f"/dev/{result.strip()}"
         if DEBUG:
             print(f"{Colors.GREEN}DEBUG: Found device from lsblk: {device}{Colors.END}")
         return device
-    
+
     if DEBUG:
-        print(f"{Colors.RED}DEBUG: Could not find device for osd.{osd_id}{Colors.END}")
+        print(f"{Colors.RED}DEBUG: Could not determine device for osd.{osd_id}{Colors.END}")
-    
     return None
 
 def get_smart_data_remote(device_path, hostname):
-    """Get SMART data from a remote host"""
+    """Get SMART data from a remote host with proper device type detection."""
     if not device_path:
         return None
 
     # Strip partition suffix
     base_device = re.sub(r'p?\d+$', '', device_path)
 
-    # Use sudo for smartctl
+    # Detect type: NVMe or SATA
-    cmd = f"sudo smartctl -a -j {base_device} 2>/dev/null"
+    if 'nvme' in base_device:
-    result = run_command(cmd, host=hostname, parse_json=True)
+        dev_type = 'nvme'
+    else:
+        dev_type = 'sat'  # sata/ata, compatible with SSD/HDD
 
+    cmd = f"sudo smartctl -a -j -d {dev_type} {base_device} 2>/dev/null"
+    result = run_command(cmd, host=hostname, parse_json=True)
+    if DEBUG and result is None:
+        print(f"{Colors.YELLOW}DEBUG: SMART data failed for {base_device} on {hostname}{Colors.END}")
     return result
 
+
 def get_device_health(osd_id, hostname):
     """Get device SMART health metrics from the appropriate host"""
     if DEBUG: