PC Locking up randomly during use

[DocGreen]

Just reconfigured one of my business client's networks (moved databases to strongest PC to act as their server, tuned all PC's using D7, created new user accts, etc)

Everything seemed to be working fine, but now I'm being told that the PC that hosts the databases is randomly locking up. It acts as if it's going into standby... the monitor goes out and gives the "no signal" OSD message and the computer stops responding to mouse/kbd input. I'm also not able to remote into it when this happens (remote command line still connects, but it doesn't execute any commands.) The only way to recover the PC is to kill the power and reboot, which could potentially corrupt the LOB databases.

To the best of my knowledge this PC has never done this before. I've disabled all of the power options (turn off monitor, turn off HDD, sleep, hibernate, etc) in Windows, but it still does it. I've verified that there are no viruses/malware hiding anywhere, checked for errors in the event log, updated drivers, but still no joy. I'm kind of at a loss here. Anyone have any suggestions before I nuke & pave the damn thing?

 

[PBComputer]

Have you checked the nic is not going to sleep, could also be a newer driver causing the issue. I would also create a new user account just in case something is corrupt. If not then n&p

Paul

 

[nlinecomputers]

So why isn't a critical LOB app on a real server? Time to explain to your client that **** like this is WHY you need to spend money on such a unit.

 

[brandonkick]

This stuff should be on a real server.


My only guess is that either the migration or tuning with D7 fumbled something up. If it worked before, and it doesn't now then most likely it was a result of any work that was done during the migration.

There is still an off chance that maybe the changed/increased work load can be triggering this issue that wasn't triggering before.

My best advice is that a database belongs on a real server. They can be purchased rather cheaply, and while the server OS isn't cheap... the migration to a proper setup will make up the cost before they know it.

 

[ohio_grad_06]

What about heat? If you are trying to run databases on a normal pc, could be having an issue there. I would back up, ruin memtest 86, if ok, I think I'd run Intel burn test to stress and great up the cpu, I use this when overclocking to test stability. May want to check what power supply is in there and that it's not doing out either.

 

[Markverhyden]

What the others said about a real server. At any rate we do not know all of the details. Personally I would not waste any time troubleshooting and do a nuke and pave. But start with a data backup as well as a complete bootable image. Making sure to test the bootable image thoroughly. Then proceed with nuke and pave. Remember that this is their server so every time it crashes and you do a hard (ungraceful) reset you risk causing problems with their LoB app. I know from experience, in spite of what various app companies claim, they cannot completely restore things. So there is a risk to the business.

 

[trevm999]

What OS? Is this a domain environment? How many PCs are connecting to it?

I agree with the others, this sounds like it belongs on a server. I would refuse to do this even if the client says that is what they want.

 

[DocGreen]

It does belong on a server, and I've explained that to them every time they call me... but for some unknown reason they'd rather pay me to come out and pick up the pieces every couple months instead of investing in proper equipment.

The PC in question is a custom build (another local shop)
• Windows 7 Pro 64-bit
• Intel Core i7 Quad-Core 3GHz
• Gigabyte H55M-USB3 mobo
• 8GB DDR3
• 500GB WD Green HD (just did a double-take on that one)
• Nvidia GeForce 9400GT
• Realtek RTL8168D/8111D Gigabit NIC


No, it's not on a domain

I sincerely doubt it's an issue of too much load on the PC... the one that was previously being used as the "server" was a crappy little eMachine's PC someone brought in from home or something. (single-core 2.4GHz, 2GB RAM, Win Vista Bus.) The database was slower, but it never had any problems.


I'm going back to this client today to cleanup a mess with their phones, so while I'm there I'll look at this PC again and see if I can't find anything funny... if not, I'll just wipe the damn thing.

 

[lcoughey]

Get a full sector-by-sector clone of the drive to a new known healthy drive (not a WD Green) and see if behaves the same way. Might save you the time of a Nuke and pave...though I'd likely not nuke and just pave a new OS on a new drive if I didn't do the clone.

 

[ohio_grad_06]

+1 on the above. I saw no mention of the PSU there, so I think I'd replace that with a known good PSU. I'm finding EVGA has 80+ certified power supplies at my local microcenter for relatively cheap recently. Get a clone to new drive, sfc scannow command, and chkdsk. Run a memtest pass or 2 on the ram. Anything in event viewer?

 

[tek9]

+1 on changing the HDD. Also consider using an Intel NIC for the "Server".

 

[ohio_grad_06]

Also a good idea. Make sure said nic is not sleeping our shutting off also.

 

[YeOldeStonecat]

Ugh...a "cloner" tower...and a "WD Green" HDD? LOL. LOL some more!

If you can't get them to purchase a true server...can you get them to purchase a little tier-1 business grade workstation at least? Spend 650 bucks on something decent? A couple of calls, interruption in work, and hours of your time....troubleshooting this cloner with the wrong HDD for a server..and they may as well have gone out and bought a new workstation!

 

[ohio_grad_06]

I can understand, though if the home brew systems are done right they can be decent. At my day job we've got one that has dual xeon cpus 5 or 6 drives, Used to be a server that was used like once a year at conferences, but since we needed to upgrade existing infrastructure on a smaller budget at the time, it got pressed into service full time. We changed all the drives for new ones, it would lock up if too hot, I discovered one day on troubleshooting that there was a plastic shroud inside it I guess to direct air and it sat like less than 1/8 inch above the fan for the cpus(rack mount system). So when the fans kicked in full they could not pull enough air and it would overheat. I pulled the shroud out, and solid as a rock for months(knocks on wood). Just seems like sometimes the home brew style systems need a little more tweaking. I write still check what power supply is in there, maybe run intel burn test on it to torture test the processor, if you do that though keep a temp monitoring program up because the chips usually heat up quick, but that may help show how stable that box is.

 

[KompuKare]

Ugh...a "cloner" tower...and a "WD Green" HDD? LOL. LOL some more!

Don't forget the 65nm Nvidia card which could very well eventually suffer from the infamous solder defect.

As for the OP's problem, is there nothing in the event log at all? Aside from OS issues, the BIOS could also be the problem and since I have certainly encountered what I call a corrupted CMOS settings (as in clearing CMOS fixed something despite all the settings already being on default) it might be worth clearing that.

 

[MMoore]

(New customer) Had a real HPG5 server giving me similar issues. I only had a remote session to troubleshoot at this point. Zero event viewer errors before I went over to inspect and possibly replace the psu, come to find out it had roughly 8 blown capacitors. I usually fix the caps but this thing has a lot of hours on it zero maintenance and was running server 2003 on a domain with 4 people in the office lol. Replaced the server added a Raid1 NAS and all is well. If you haven't checked take a peek for blown caps on the video card too.

 

[DocGreen]

I've got it on my bench now and have watched it go stupid on me. It sure does seem like a sleep/hibernation issue. According to the person I talked to when I picked it up, it only ever happens when its idle, which is consistent with what I'm seeing now. Tweaking the bios settings now, if that doesn't help I'll try clearing the CMOS as suggested.

I'll also double check the event logs and caps.

 

[DocGreen]

Wiped and reloaded... problem reappeared immediately after booting.

Pulled the video card just to test against a known-good card and noticed that the GPU felt like it was on fire. Would an overheating PCIe video card cause the whole system to hang?

 

[bg_bls]

Wiped and reloaded... problem reappeared immediately after booting.

Pulled the video card just to test against a known-good card and noticed that the GPU felt like it was on fire. Would an overheating PCIe video card cause the whole system to hang?

I had a system about 2 weeks ago with an overheating video card that would freeze up the whole computer. It happened more often during sleep, but also while using it too.

 

[ohio_grad_06]

Yep. Maybe look for something cheapish that has decent passive cooling and toss a case fan in next to it and call it a day.