Question Home Server no longer boots after Power Outage ?

[Repareman]

So I have a fair bit of information that I'd like to share and hope to get some help with this. I have build a home server which I do a number of various things with using Unraid. I have an Epyc 7601 on an AsRock RomeD8-2T motherboard with a Be Quiet! Dark Power Pro 12 1200W power supply.

In the morning, the power at home went out. At this moment, I was currently monitoring my server through a power meter directly into the wall outlet. So no surge protector. (I know, bad.) While the power was still out, I actually unhooked this, and plugged the power cord to the server into a UPS that also offers surge protection. So when the power came back on, the server was behind a surge protector. To my understanding, when the power comes back on is when the surge would occur.

Following all of this, when I noticed the power was back on, my server would no longer turn back on, as if power was not being delivered. Which leads me to debugging for the last day. Allow me to share the symptoms and some of the odd behavior that I've noticed.

With the power supply that I currently have, the odd behavior I noticed was the green light on the motherboard that used to always be green constant. It was now powering on to brightest green, then fading away like it suddenly lost power, and then continuously doing that just over and over again. So I pulled the 24-pin from the motherboard and I jumped it. This caused all the fans and hard drives to start spinning again like power deliver was working just fine. Not satisfied that this completely absolved my power supply from failure, so I grabbed another 850w power supply that I know to work. (I double checked it on another computer and worked fine) When I plug in the known working PSU and just turn the switch on the PSU to allow power to flow, I have not hit the power button on the server yet, it tries to power up, spins the fans on the PSU and the case a little bit for half a second, then stops. Then tries again, and stops. Just power cycling constantly without actually turning the computer on.

I have removed everything from being hooked up to the point where it was -ONLY- the CPU, Motherboard, and tried both PSUs. I get the same behavior. Also tried with and without RAM. I've tried clearing the CMOS. I've removed the CMOS battery for an hour and hit the CMOS clear jumper like a madman multiple times and the same behavior. I see no damaged capacitors on the motherboard. I've ordered a PSU tester to make sure that voltage is being delivered in a clean and reliable manner, so I can try to update this posting with results tomorrow when I get the tester. I'm really really hoping that someone knows like some other things I can try to check for. I really don't want to have to buy a whole new server motherboard and Epyc CPU...they're not cheap....this thing is like a couple months old....sad day.

 

[Ralston18]

Probably not relevant per se but just for the record - what make and model power meter?

More relevant though is the UPS: make, model, age, specs?

Was the UPS connected to a wall outlet and was the battery charged before the server was plugged into the UPS?

What surge protection rating (joules) for the UPS?

FYI:

Surge Protection and Joules

The concern being that the UPS may not have actually offered any surge protection. Protection having been lost due to some cumulative number of smaller hits or a few larger hits.

Also: any power strips, other surge protectors anywhere along the power path to the UPS and server? Daisy chaining (connecting the devices in a line of some sort) is not recommended. Plus some devices have small built-in circuit breakers that can trip and thus no longer provide power to hosted devices.

When the incoming power is stable try connecting the server directly to a known working wall outlet. Determine if the server receives power.

If so, then try again with the UPS in between: Wall outlet ----> UPS ----> Server. Determine if power is received.

Objective being to discover where in the connection path that power is lost or becomes a problem of some sort.

 

[Repareman]

Probably not relevant per se but just for the record - what make and model power meter?

More relevant though is the UPS: make, model, age, specs?

Was the UPS connected to a wall outlet and was the battery charged before the server was plugged into the UPS?

What surge protection rating (joules) for the UPS?

FYI:

Surge Protection and Joules

The concern being that the UPS may not have actually offered any surge protection. Protection having been lost due to some cumulative number of smaller hits or a few larger hits.

Also: any power strips, other surge protectors anywhere along the power path to the UPS and server? Daisy chaining (connecting the devices in a line of some sort) is not recommended. Plus some devices have small built-in circuit breakers that can trip and thus no longer provide power to hosted devices.

When the incoming power is stable try connecting the server directly to a known working wall outlet. Determine if the server receives power.

If so, then try again with the UPS in between: Wall outlet ----> UPS ----> Server. Determine if power is received.

Objective being to discover where in the connection path that power is lost or becomes a problem of some sort.

Hey Ralson! Thank you for replying for more information!

So the power meter that I was using was this: https://www.amazon.com/gp/product/B07DPJ3RGB/ref=ppx_yo_dt_b_search_asin_title?ie=UTF8&psc=1
I don't believe it offers any surge protection at all, just overload protection. So when the power went out, it was setup as: OUTLET --> Power Meter --> Short Extension Cord with a ground --> Server
When the power came back on, it was: OUTLET --> Power Meter --> Short Extnsion Cord with a Ground --> UPS --> Server

The UPS is the CyberPower LX1325GU (https://www.cyberpowersystems.com/product/ups/battery-backup/lx1325gu/) Says that it should protect Surge Suppression in (Joules): 890 Joules.

The UPS is brand new. Like the sad part and to further show off my stupidity in this situation is that it's about a week old from Micro Center. My server was doing a couple things and I figured, "I haven't had a power outage here in the 4 years I've been here, what are the odds it will happen by next week." And here I am. I had already charged its battery. However, I will say that the UPS was not turned on when the power came back on. So the battery wasn't running, but I didn't think that it would need to be turned on for the surge protection to work? I could be wrong here.

I tried to connect the server to a known working outlet with the same behavior.

I am also going to try to get a multi-meter and start checking the motherboard today with an online tutorial and try to update my post with more information with what I find.

 

[Ralston18]

Is the UPS still in power path?

If you have not done so, go back to the original power path without the UPS being connected.

Just to eliminate the PSU as a possible culprit. Check its' circuit breaker.

= = = =

Testing:

Start with the PSU:

https://www.lifewire.com/how-to-manually-test-a-power-supply-with-a-multimeter-2626158

Not a full test because the PSU is not under load. However, any voltages out of tolerance make the PSU a suspect.

Also: "start checking the motherboard today with an online tutorial "

What online tutorial are you planning to use? Link?

 

[Repareman]

I'm attempting to follow guides similar to this: https://amazeinvent.com/test-a-pc-motherboard-with-a-multimeter/

But not having much luck in terms of following the steps, still trying to work my way through it.

I have tried without the UPS and gone back to the original configuration as well, same issues continue to persist.

I'm not sure how to check the PSUs circuit breaker. And I have tried another PSU that I feel is known to work. So if server PSU is PSU-A and second PSU is PSU-B, then PSU-B works on a different computer that can boot up fine, it's 850W, which should be enough to turn just the motherboard and CPU on I feel. And PSU-B causes fans to try to spit for a split second before shutting off and going into a power cycle. PSU-A just causes the stand-by green light on the motherboard to light up and then fade in it's more minor power cycle loop. (I hope that made sense.)

 

[Repareman]

Additional note. So I have my PSU plugged into the board, has power. And what I'm seeing when I put my multi-meter on the motherboard chassis and the positive end on the positive connector where the Power Switch would be plugged into the board, it seems like the voltage is not a constant, it is going up and down. Which I think goes in line with the green stand-by light powering up and down constantly.

Sigh This is really sounding like a dead motherboard. But would love additional thoughts if someone thinks that it might be something else.

 

[Repareman]

I feel implored to let everyone know the outcome here. If by happenstance someone stumbles across this.

So I'm an idiot. When my power went out, I took the outage as an opportunity to hook up my new UPS to my server. After all this happened and I couldn't figure it out, I actually ordered a new mother board. I move everything over, swapped it all out and it tested just fine. After I put everything back in and attached everything, it was experiencing the exact same problem. It took me about 30 seconds to pull the USB cord from the motherboard that went to the UPS and then it all worked again. I moved everything back to my original board that I thought was dead and without that USB it worked just fine. I did not follow Ralson's advice well enough as I should have. I switched power supplies and outlets and everything, but I never thought to pull out that USB cord. I never thought that a USB could do something like this. I actually tried with a different USB cord and it worked. So -ALL- of this was just a bad USB cord.

I hope other's follow instructions better in the future when people tell you to unplug things.