UNSOLVED
Random UPS Shutdown During Memory Test (LCD error PWR1006 when running Windows Server)
Hi All,
We are low on budget and just went through massive layoff. We have an ongoing project to to add 10 new servers to replacing existing old R510 ones.
Having no budgets, we bought 10 used R620 off eBay resellers. Each has dual XEON E5-2680 2.7Ghz, with total of 16 cores/32 threads, 64GB memory and 8 brand new Dell certified 600GB 10K drives. These R620 servers have no Dell warranty.
We updated all firmware to the latest for all servers. Each comes with dual redundant 750-watt PSU. Each servers connect to two different UPS for redundancy. We have been using these two UPS for about 4 years without problems (for this rack).
Here is what we found.
1) Ran Dell memory test. Under 2 weeks, both UPS were shut off in the middle of the night. We knew we got lemons from eBay resellers. The question is which one(s). I have no ideas which one(s) caused both UPS to be shut off.
2) Because the rack was also used by production servers, we had to buy two brand new UPS to test these on a different rack.
3) Now, on a different rack with two brand new UPS. It doesn't shut off the brand new UPS this time. The newer UPS may be able to handle the power problems better.
4) We swapped the UPS to test the used R620 with the two older UPS.
5) No shut down yet with the same memory tests.
6) We installed Windows Server 2012 and the LCD has errors "PWR1006 System power exceeds capacity. System halted. Contact Support". It shuts down Windows Server.
It seems the problem may be the 750-watt PSU. I checked Dell doc at https://tinyurl.com/y49vsr58 and found each R620 most likely requires the 1100W PSU. We are not sure if the hardware itself is suitable to be repaired and be deployed for production.
My question is because we have been running hardware test for months using these under-capacity PSU, will it cause random hardware problems in the long run? I heard under-capacity PSU could damage logic board, CPU, etc (specially if used for a long while). We might have to cut our loss and buy new servers.
Let me know what you think.
Chris
