Win 2012R2 stops responding for a few seconds

[maniek765]

Without any obvious cause Windows 2012R2 just freezes for 3-5 seconds, then it resumes until next freeze.

I've reinstalled windows, each time ran HP Service Pack for Proliant for correct drivers. Each time only failed package is ROM update for HP Smart Array RAID controller. I suspect this might be the cause. In the package install log however there is no usable information.

Also, I tried running Windows 2012 (not R2) - this runs fine.

[waaronb]

What model Proliant is this?

When you say it "freezes", are you logged into it over the network and it freezes the display for you, or are you logged in locally and nothing happens for a few seconds?  I'm trying to get you to think about if it's a network glitch or an actual server issue.

Run some performance logs on there and see if there are periods of high CPU usage.  If the machine totally froze up but then was fine a few seconds later, it's probably software related.

If there were any hardware problems like a hard drive or memory issue, it should have been logged in the IML so check that out to make sure it's all good.

[maniek765]

It's a Proliant ML310e Gen8 v2

I'm working on it locally so it is not a network issue. Also, most of the processes also notice that "freeze", for example:

- I open up Task Manager and go to performance->ethernet

- I start downloading a big file from LAN

- Ethernet throughput is at 70% and steady

- A freeze happens

- after the freeze chart updates itself and I can see that there was almost no transfer during freeze time

- everything goes back to normal until the next freeze

Actually I traced down the problem but found no solution. As long as I'm NOT using the embedded Smart RAID controller - I disable it in BIOS and go for simple SATA and non-raid disk config, everything is fine.

I updated the RAID controller drivers, run firmware update through Intelligent Provisioning, even installed all possible Windows Updadates. No effect.

[waaronb]

I'd suggest running performance monitor and track your disk activity and CPU.

It sounds like your problem isn't that the whole system is freezing, but the file transfer is freezing for a little bit? That would be different... and that could be because it's buffering the incoming file and then it flushes it all to disk. If the disk isn't fast enough to keep up, then it'll pause the transfer.

Looking in performance monitor at all of the disk-related items would help you narrow down the problem. In the logical disk section you'd want to monitor the queue lengths, read/write bytes per second, etc. And also add in the network read/write bytes/second so you can see it all together.

Maybe you could describe how your disks are setup? Does the RAID controller have any write-cache, and how many disks are in the array? Is it RAID 0, 1, 5, etc?

Just based on your symptoms, my first guess is that the disk subsystem is simply not able to keep up with the amount of data coming in from the LAN. If you had a gigabit LAN that was moving data pretty well, a single non-RAID disk like a slower RPM SATA or SAS drive probably can't keep up. If the file is big enough, at some point Windows needs to flush the data to disk if it even has write-caching enabled at the OS level.

If your RAID controller doesn't have any write-cache or it's setup to read-only caching, that's another area where it would cause some "stuttering" during large file copies.

Another interesting thing to look at is how you're copying the file? Drag and drop from Explorer, or command line via copy/xcopy/robocopy? FTP maybe?

You will get wildly different performance from each of those... if I'm moving large files I'm going to be doing it via command line and robocopy or xcopy, maybe using the /J parameter (although honestly I don't notice much difference in Server 2012 with or without /J)

Explorer's file copies in Server 2012 are a lot better than they used to be, but they still seem a LOT slower to me than what you'll get from a command line method.

I'd guess your CPU probably isn't too busy during the file copy, but I'd monitor it too, just in case... maybe you have a virus scanner or something getting in the way.

And if any of those disk queue lengths during the file copy start climbing higher and higher (like the write queue) and then you notice the LAN activity stop while it clears out that queue, then you'll know that's what's happening.

My prediction is, you'll see the write queue get higher and then at some point the net activity will drop while the write queue clears up... you should be able to see how the disk write bytes/second compares to the incoming network traffic and if there's an imbalance.

[maniek765]

The case of network file copy was just an example how a "freeze" looks. They happened regardless of network or heavy disk operations.

Now everything is running fine, I got it working by changing both hard disks.

I'm very suprised that almost identical (MB1000GCEHH and MB1000GCEEK) drives won't work together. What if a drive fails and I wont be able to get an exact replacement? I'll have to buy two new ones?