Untangle router losing power

[tek9]

I recently replaced a failed Untangle custom built router at a customer with a new custom build from MITXPC.
The old router had a power surge knock out their WAN port a while back. I switched their WAN port to an unused one. After that a few more power outages basically killed the unit so I replaced it with a custom built unit from MITXPC (this model: https://mitxpc.com/products/nml-nf692), one week ago.
Since then, they've been having issues with the router experiencing a power loss and ungraceful shutdown randomly a few times a day. The router boots up pretty quickly on its own after that.
They haven't had any power outages during this time, and the Unifi devices I have on this network are showing consistent uptime so they're not going down. It's just this router. I though perhaps the power isn't "clean" enough or something so I installed an APC Back-UPS BX1000M unit there this morning. I just logged into the router and see that it rebooted around 2 hours ago so that didn't help.
I have these same exact routers deployed at two other clients and they haven't had any issues.
I'm thinking it might be the unit itself or the power brick that's causing this.
How can I test this before replacing it again, and does anyone have any input as to whether it might be a different issue and I'm looking in the wrong place?
Thanks in advance.

 

[Markverhyden]

what does /var/logs show?

 

[YeOldeStonecat]

I'd be thinking it's the unit itself.....or its power supply. A little low on price so supposed I'd question the quality.

 

[tek9]

what does /var/logs show?

As far as I know Untangle does not support cli use, but I opened a ticket with their support team and they sent me a screenshot of the cli and said that it showed a power issue due to an ungraceful shutdown.

 

[Metanis]

I note it is a fanless unit. Is it overheating?

 

[tek9]

I note it is a fanless unit. Is it overheating?

I don't think so. It's in a ventilated closet and doesn't feel hot. Besides, it turns right back on within a second or two and stays on for a few hours.

 

[CLC]

Is it really losing power or just rebooting?

 

[Markverhyden]

As far as I know Untangle does not support cli use, but I opened a ticket with their support team and they sent me a screenshot of the cli and said that it showed a power issue due to an ungraceful shutdown.

http://wiki.untangle.com/index.php/Enable_SSH

Use root for the username and the password for the admin account. BTW this opens it for ALL interfaces, including WAN.

 

[Sky-Knight]

Ignore the wiki, make your own TCP 22 access rule that isn't stupidly exposed to the world. Heck simply adding source interface non-wan is a huge help.

Using that default rule gets you hacked... every time.

 

[tek9]

Update:
So I ended up replacing that router with a similar unit. Brand new out of the box, installed a stick of Crucial 8GB RAM and a Samsung Evo 860 250GB SSD. Installed Untangle, restored settings from a backup and set it up at the client.
It worked for about 10 days, then stopped. It didn't lose power or restart as the last one. The lights are still on/flashing etc. but the network loses Internet and I am unable to login to the web gui. Only way to get it back up is to pull the plug. Once done, it boots up within a minute and is fine for a day or two, then we do it all over again.
Opened another ticket with Untangle support but they're blaming hardware.
My question is what would cause the unit to lockup at random? I have similar units with the exact same parts elsewhere but they work non stop without issue.

 

[Computer Bloke]

When you try to get to the GUI are you connecting from within the LAN or from outside?

And is your Untangle box auto-negotiating link speed on the WAN port or do you have it set manually?

I'm wondering whether a transient failure of whatever device the WAN port is connected to (let's call it a modem, though it might not be) could be causing the Untangle machine to fail to renegotiate the WAN port speed. You could try popping an unmanaged switch in between the WAN port and the modem to see what happens - it might tell you something, or it might just make the problems go away.

In any case, all the network components need to be on a UPS, not just the Untangle machine. Some devices can be much more sensitive to power fluctuations than your average office computer and you might think there were no power outages while living in Brownout City.

 

[tek9]

When you try to get to the GUI are you connecting from within the LAN or from outside?

And is your Untangle box auto-negotiating link speed on the WAN port or do you have it set manually?

I'm wondering whether a transient failure of whatever device the WAN port is connected to (let's call it a modem, though it might not be) could be causing the Untangle machine to fail to renegotiate the WAN port speed. You could try popping an unmanaged switch in between the WAN port and the modem to see what happens - it might tell you something, or it might just make the problems go away.

In any case, all the network components need to be on a UPS, not just the Untangle machine. Some devices can be much more sensitive to power fluctuations than your average office computer and you might think there were no power outages while living in Brownout City.

Connecting from within the LAN.
It's a static IP on the WAN side, but link speed is auto.
Everything in that closet in connected to a UPS. Most are connected to a surge protector, but the router is directly connected to the UPS.

 

[Computer Bloke]

My question is what would cause the unit to lockup at random?

It sounds as if you've done everything you can to eliminate the power supply as a source of the problem. I'd still be wondering about overheating; even though the device's case might not be particularly hot you can't be sure what's happening inside. Do you have any way of monitoring the device's CPU temperature directly?

And of course the other question is "are you sure it's really locking up"? I know it's a headless machine, but it has USB ports - can you get a USB video adapter and keyboard/mouse on there so that when it fails you can see what's happening independently of the network ports?