Neal Spangler

Work Endpoints

Two failing shop PCs diagnosed while work kept moving

The RIP station kept hanging and the shipping PC kept blue-screening. I tested both, added capped smart-plug recovery for the RIP station and worked through Windows power settings and a leaking graphics-driver helper.

Windows production PC
Windows production PC Photo: BrokenSphere, CC BY-SA 3.0
112+ hwithout a hang after the Windows fixes
42,644 → 195open handles on the leaking driver helper
~21 minfirst recovery with nobody on site

The problem

The RIP station hung every one to three days, left nothing useful in the Windows logs, and held up print jobs each time. Meanwhile the shipping PC, which runs the carrier software, kept blue-screening.

What I did

  • Added a heartbeat log. One restart showed a clean shutdown, then over an hour with no Windows running, which put that fault below the operating system.
  • Tested memory, firmware, the power supply and CMOS battery one at a time, dropping theories when the results did not support them.
  • Put the station on a smart plug that power-cycles it after a confirmed hang, with a daily cap and an alert when a person is needed.
  • Changed Windows power-saving settings, including the plan the RIP vendor requires, then found a graphics-driver helper leaking handles and cleared it without a reboot.
  • Backed up the shipping PC and verified the copy before touching it, then worked through its blue screens with memory tests, system-file repair and crash-dump triage.
Handle leak before and after Bar chart on a linear scale. A leaking graphics-driver helper on the RIP station held 42,644 open handles. After the leaked process was cleared, the fresh one held 195, a bar barely wider than a line. 0 10,000 20,000 30,000 40,000 open handles (linear scale) Before 42,644 After 195 cleared without a reboot Open handles on one driver helper
Handles held by one leaking driver helper, before and after clearing it. Linear scale.

What I took from it

The logs gave me little to work with, and my hardware theory did not explain the improvement after Windows power-setting changes. I kept testing rather than calling the fault resolved.

Privacy