Your WMI sensor is red and the error looks technical. It almost never is. WMI monitoring stands or falls on one account, its password, and its permissions - here is the whole story.
CPU and RAM numbers in PRTG usually come from WMI - Windows Management Instrumentation, the built-in channel Windows offers for exactly this data. When a WMI sensor fails, beginners reinstall things. Experienced admins check three things: the account, the password, and the firewall between probe and target. This page walks the credential reality first, then the honest cost of WMI, because WMI's power and WMI's weight are the same feature.
Create a dedicated account for monitoring - in a domain, one service account; in a workgroup, the same username and password on each machine. It needs local administrator rights on every Windows target it queries, because WMI performance counters sit behind admin boundaries. Enter the credentials once at the group or device level in PRTG and let sensors inherit them. Using your own personal admin account works until the day you change your password and every sensor in the office turns red at once - that day teaches this lesson permanently.
WMI rides on DCOM and RPC: the initial connection uses TCP port 135, and the actual conversation continues on dynamically assigned ports. On the same LAN segment with the Windows Firewall's WMI rules enabled, it simply works. Across a firewall or a routed network, that dynamic port behaviour is why WMI "mysteriously" fails - the fix is the proper firewall rule set for WMI, not disabling the firewall. Test from the probe machine: if the probe cannot reach the target's WMI, no sensor setting will save you.
WMI queries are not free. Each one is a real conversation with the target machine, and Paessler's own planning guidance caps a probe at roughly 120 WMI sensors at a 60-second interval (about 600 at 300 seconds) before the probe itself becomes the bottleneck. For a handful of servers this is irrelevant. For fifty servers, prefer longer intervals on memory and disk, and take switches and network gear by SNMP instead - SNMP is the light protocol, WMI is the rich one. Rich and heavy are the same thing, priced differently.
Nine times out of ten the ladder ends at step 1 or 2.
Every sensor in this guide is created live in the video - installation, discovery, CPU and RAM, bandwidth and firewall status.
Watch Full VideoNot through WMI - the performance data sits behind administrator boundaries in Windows. Alternatives exist for some data (SNMP on servers that run an SNMP service, or performance counters exposed other ways), but the standard, reliable route is a proper admin monitoring account.
Shared cause, shared failure: the monitoring account's password expired or changed, the probe lost network access, or a firewall rule changed. Simultaneous failure is a gift - it points at the one thing all sensors share.
They answer different questions. SNMP on Windows is limited and needs the SNMP service installed; WMI gives full CPU, memory, disk and service detail. Use WMI for Windows depth, SNMP for network devices, and respect WMI's weight on the probe.