Busy with the family yesterday but I ported my monitor program to a raspberrypi and set it running. Engineering reason: raspberryPi’s use much less electricity than a PC or laptop and can be left running headless (no screen or keyboard) in a corner, continuously.
Every two minutes it measures how long ‘the forum’ takes to respond to 7 different adverts, plus a delay of 12 seconds between each individual call, so the advert times are taken once every 204 seconds. Simulates a fast-ish human typist tirelessly browsing the forum with a stopwatch and it takes no comfort breaks. And is super-human in that it records the time taken by internal calls, not just the whole page. Engineering reason for measuring internal calls is that it helps pin down what’s amiss rather than simply noticing “the forum is slow”.
As I type this the forum appears to be in good order, responding quickly, though AdBlocker is ON. My individual observation doesn’t mean much because whatever is slowing the forum down is intermittent. It was OK for me Sunday morning, and again when I logged in briefly before bedtime. All fixed then?
This morning’s graph suggests not!

- No problem between 1700 and 0100
- Just after 0100, there are 4 separate major slow-downs, with brief gaps, making the forum unusable for 5 hours
- A few minutes after 0600 today, just before I logged in, the forum started responding normally. Long may it last.
The first run of this program strongly suggested the classicmagazine php adverts were causal. The latest graph shows all advintage adverts are affected. The Amadeal advert normally vends in less than 0.2s, last night it was taking 10s, an x50 hit. The already slow classic mags group normally taking about 2s average 10x delay, and one of them takes over 36 seconds to arrive.
Last night’s performance figures are dreadful. Cause unknown to me. Possibly overnight maintenance, not necessarily human.
At work, had an incident like this, always out of UK normal hours. After a long-winded technical analysis failed to find anything wrong, eventually twigged the problem was due a cleaner working an irregular schedule. This lady had a key to the secure computer room and had got into the habit of unplugging a big network router so she could use the socket for her vacuum.
The socket was marked “ELECTRONICS ONLY” in red, but her common sense prevailed and of course the vacuum cleaner worked perfectly. To her that red warning was officious nonsense. The big network router was backed up by a smaller one with a slow network connection. The system didn’t fail, but the slow connection caused massive queues. Cleaning done, the our lady plugged the router back in and 20 minutes later the queues were cleared. This system ran 24×7, busiest during the UK day, but also supporting shift work and many international connections.
Lesson learned: brief the cleaners, taking care to explain they must follow the process, observe notices, and above all not apply ‘common sense‘. Turned out that the cleaner had an extension lead she was supposed to plug into the utility socket provided. But so much easier to unplug a mission critical router from the clean UPS protected circuit, taking the router out of service and squirting electric motor muck into other sensitive equipment.
No medals issued for sorting this out. The powers that be said it had taken the techies far too long to identify a simple cause. Engineers always get the blame!
Dave