Poor Design Wastes 3 Slots In a PC Case 6
Ever wonder what PC card designers are thinking? A system here has an adapter that converts Infiniband into 4 internal SATA ports. That’s fantastic, but the design of that adapter is less than desireable. It takes up 3 PCI slots. I’ve lost the use of 3 PCI slots just due to poor design.
See the card with 4 SATA cables at a 90 deg angle? See how it gets bent into the slot above and how the cables press against the NIC below it? All the slots below the SATA cables are filled, so moving it to the bottom slot isn’t an option, sadly.
It has always been an issue and I knew it was, but now I need a PCIe-x1 slot for a USB3 adapter and can’t install it because of that crappy SATA converter.
In order to use the new card, it appears a NIC will have to be removed. I’m not thrilled about that. I’ll remove the NIC in the last slot, so hopefully this will open another PCIe-x1 slot for use too. Hopefully.
By the time you read this, I will have taken the system and all the VMs down (including this blog) and done the changes.
So that was fun, not!
NIC Changes
When you take a NIC out of a system under debian/ubuntu, strange things happen to the network device names. In theory, I’d addressed that in the /etc/udev/ files, but when it didn’t work out that way, that’s were my troubleshooting began.
I had to relearn
Seems having a default route set to a NIC that doesn’t exist anymore isn’t good. Who’d a thunk it?
Since about 5 VMs automatically startup when the hostOS is rebooted, shutting down the box after making changes was a hassle too. VM hosts aren’t meant to be rebooted that often.
Looks to be Fast
Still, having a 2TB USB3 disk and room for another is a bonus. So far, write performance looks to be 2x my SATA1.5 devices, but only a mkfs -t ext4 is running. Just to be clear, I did split the disk in half via partitioning. I expect to drop some non-production virtual machines on this new disk.
Automounter
I need to get automounter managing this disk since it is a WD USB Green that is known to shutdown when not in use.
PSU Fan Issue
While I was behind the rack working, I noticed that a PSU fan wasn’t spinning on one of the other machines. It has 2 extra fans pulling air out and the temperature from both exhaust fans isn’t warm at all.
I haven’t been inside that case in about 3 yrs. It is the Xen server running most of the production VMs for 2 companies. I probably need a pilgrimage to Microcenter today. I was across town at Fry’s looking for a PSU for a completely different reason yesterday and carried the Microcenter ad with me. Fry’s didn’t have any brand-name PSU deals. I need to check the purchase date of the PSU in that machine. I think it is an Antec 80+ that I already RMA’d after 2 yrs of use. Time to test my paperwork filling system … less than 20 seconds later:
It is still running, so I’m hardly mad. It is quiet – and has always been quiet.
I love running computers. Really, I do.
Some days you are too smart for your own good. For me, today was one of those times.
With the extra storage connected, I decided to move some ISO files out of my HOME onto the new storage. Simple enough, right?
Nope.
This server runs KVM. It has about 5 VMs on it. All of them started up fine, except 1. The one that failed to start was my wonderful Windows7 Ultimate that records TV shows.
All the other VMs were linux-based. This is the only non-Linux VM. What’s different besides that? I look at the syslog file for a hint.
Gee, that isn’t helpful at all. Ok, my google-fu found a few bugzilla pages for libvirt and the code fix was some libvirt C code. More than I wanted to deal with. I had hoped it was just a KVM python script that needed fixing. Oh well. Inside virt-manager, the Windows VM settings wouldn’t come up. Again the error message was less than helpful. Back to syslog and into the libvirt.log for the VM.
Ah – something helpful.
The virtio ISO file that was mounted under the CDROM device for the Win7 VM was moved. The VM can’t find a CDROM image to connect. That’s the problem preventing the VM to boot. Seems dumb to me – whatever. Perhaps a newer version doesn’t have this issue.
Ok, so if I remove that setting in the XML definition file for the VM, life should be good – right? Nope. Seems the VM XML is “defined” – I guess that means the VM settings is cached somewhere else on the system. The fix:
Good enough.
The VM boots now and I can solve other issues after wasting at least 45 minutes on this one.
Complications.
Space problems? It’s usual now that GPUs are getting bigger all the time. http://dl.dropbox.com/u/11392968/gpu.png Do you see the pci port? Most probably no because the second GPU almost covers it: http://dl.dropbox.com/u/11392968/hidden_pci.png
Right now I don’t need anything PCI but when I had something to put there it was a pain and it blocked the card’s fan. Not to mention it was hard to find a motherboard to actually have a PCI port and being capable of SLI at the same time.
The GPU in that photo above is my most powerful and largest. It blows away everything else pushing the server console (80×25) text!
If I had 2x nVidia GTX460s, I’d be cracking passwords FAST!
Interested in getting started? GPU-based cracking.
I know how you feel, I can feel the speed when working on a tty!
I will look into cracking, it’s on my to-do list along with other things I want to learn this year. These cards definitely have some serious firepower, maybe I should start encrypting some stuff so I can crack them later.
So I went to swap out that PSU yesterday and found the fan was spinning. Guess it only runs when the heat makes it necessary. That’s a good thing to keep noise down. I feel a little dumb for not recognizing this. The exhaust fan on the center-top of the case never pushes warm air, so I should have realized cooling wasn’t an issue in that machine.
That machine is relatively busy. It runs Ubuntu 8.04 Server with about 5 Xen paravirtual VMs – including Zimbra. In looking at the yearly performance statistic graphs (a few summer months are missing), it is clear the machine doesn’t break a sweat. CPU, RAM, disk storage, disk I/O, network I/O are all fairly low. There is no swap being used at all – ever. RAM use is only 5GB out of 8GB available.
Now I have an extra PSU. That’s a good thing, though none of mine have failed in since 2/2010. Prior to that, it was 2006.