Wierd power issue & issues troubleshooting

[aconnell]

A few years back I build a custom server that serves the role as a silent virtualization system (details w/part list here: http://www.andrewconnell.com/blog/articles/creatingcustomquietvirtualizationrigwithhyperv.aspx ). I've always had a problem with the power. When I hit the power button, sometimes it come on, but most of the time it doesn't... eventually I could get it to boot up. Recently I've been able to reproduce it 100%:

- Scenario: server is running and I shut it down
- Once shut down, if I push the power button to turn it back on, it doesn't respond
- Flip the PSU power switch off for 5-10 seconds (confirming power has been killed as the NIC lights go out)
- Flip the PSU power switch on... wait 5-10 seconds
- Either the machine will turn on by itself within 5-10 seconds or I can turn it on using the power switch or remote web-based iKVM software
- Everything runs fine until I shut it down again

I'm not sure how to troubleshoot this... my guess is that the PSU is possibly underpowered, but I can't recall a time where the server has shut itself off due to what would seem a underpowered PSU. Earlier this year I built another machine very similar to it (parts here: http://www.andrewconnell.com/blog/archive/2011/07/14/updated-silent-virtualization-server-specsndashperfect-for-a-home-office.aspx ) and it has none of these problems. One big difference though, the promblem server has a 1000W PSU while the new server has a 1300W PSU. Granted the newer machine also has more HDD's in it...

Any ideas on how I could troubleshoot this? I've looked in the system logs from the board and can't find anything that stands out. The relevant parts involved are an ASUS DSEB-DG board, Thermaltake Toughpower 1000W & Cooler Master Cosmos 1000 case.

 

[wathman]

I can't see your hardware specs on your links so I can only try to guess at what the possible problems are. your PSU might be underpowered, though you have to look at the overall power profile of your system to determine that. A server with lots of drives is going to demand a different power distribution than a gaming PC that needs lots of current on the 12v rail to feed the GPUs.

If possible, can you disconnect all non-essential drives and peripherals and see if you can get a normal boot that way? Also, you don't mention how old this system is exactly. Every year a PC ages, the PSU looses some capacity to reach peak power as it did when new. Degradation can also be even worse for PSU's in servers and from the description it sounds like the ones you are using are gaming/enthusiast grade and not quite enterprise level. It could be that the PSU is just reaching the end of it's useful life after constant use for so long.

 

[aconnell]

Looks like the forum software garbled my URLs... sorry... it added a ")." to the end of them. They work now after adding a space.

While I hear you on the age of the system, I've had this problem since day 1. There aren't many peripherals on this server, just a mother board (granted dual XEONs), 4 HDDs, lots of RAM and a DVD drive... no graphics cards aside from the onboard mobo one. It's a headless machine... no input devices... just network cables & power plugged in (I connect to it via RDP & a web admin interface provided by the mobo).

I guess I could disconnect all the HDD's and try to turn it on... that would eliminate a major power draw and it would give me the instant on responsiveness a power button should give you I'll report back after that test...

 

[wathman]

I looked at the 2 PSU's you have listed and ended up scratching my head. The output current per rail is nearly identical, though the old one is listed as 1000W and the new one is listed as 1200W. I was thinking that maybe the new PSU was supplying more amps to one of your rails for your new system but I'm not seeing it... very confusing manufacturer math going on.