Replace.IT – FSP Aurum Pro 1000W Power Supply with Corsair HX650W

The last day and a half was quiet as our internet had been capped until this morning. During that time I have been having a struggle with the FSP Aurum Pro 1000W power supply unit that I installed into my new VMware ESXi 5.5U2 server.  I was using the VMware OVF Tool to export virtual machines from my old server to the new server and large machines would take a long time to copy – so I left it running overnight.  I was finding that the new server would be off.

It seems that after running for a time the power supply shuts down.  I remembered that I had faced this problem before during my cryptomining phase, but now I am not running that sort of power – my power meter indicates that power consumption of the new server is only 120W – low enough that the power supply should be able to handle it with ease. Unfortunately, that is not the case – it kept tripping out.  I know that it has tripped because the power button on the case does nothing until the switch on the power supply is turned off then turned on after a few seconds.

This power supply had been working so what has changed.  Then I had a hunch – the power supply also has two dedicated fan connectors for attaching fans – I do have a couple of fans on one of those cables – could they be the cause of the problem.

The Antec 1100 case has a fan power hub, that takes a molex connection and spreads it out to four 3-pin fan power connectors allowing additional case fans to be run.  I disconnected those fan cables and the power supply started running consistently.  That is until I thought I would test whether the fan cable can handle just one fan – I connected a single fan while everything was running and immediately – everything off.

Bad move!  After doing this, no matter what I did, removed all fans, removed the modular cables going to the dvd drive and the hard disks, leaving only the motherboard connected – the power supply was not going to work this time.  After resetting the power supply, press the case power button – the cpu fan would start spinning, then everything stops.  It happened more than 10 times in a row last night so I was ready to just throw the FSP power supply into the pool.  After calming down a little – I decided to “replace.it” with my Corsair HX650W power supply that I have in my Compaq Presario desktop.

I got out the original Compaq power supply from the Presario that I had put into the Corsair box – it was a 300W unit.  I had bought the replacement as I had previously upgraded to a nVidia 8800GT graphics card that needed more power, but have seen gone to an AMD Radeon 7850 which uses less power.  I removed the Corsair power supply – ever notice how much dust computers accumulate?  Also noticed a blown capacitor on the motherboard – that might explain the occasional blue screen that I had been getting recently – another story.

Now with the Corsair HX650W installed in the server – it powers up and I am happy to continue with my exporting and importing of virtual machines.

One thing that I do think of, is because of the multiple times that VMware has crashed due to the power supply shutdown – some of the data on the hard disk could be corrupted.  I don’t have a battery backup for the Adaptec 5805 array controller so any data that is going to the hard disk will not be stored.  It might be a good idea to buy the battery – it is about $150 or so and would allow data to be stored in the array controller cache then written to the disks on the next power up.  Also the Corsair power supply being a professional series – has a 7 year warranty, I will need to find the receipt and keep it handy, just in case.

[Note] Due to the multiple unintentional shutdowns, it might also be a good idea to reinstall the VMware ESXi to ensure that all the data is valid – I might do this after I get the battery and enable all the write caching, now that I have a good reliable power supply in the server.

Retask.IT, Replace.IT – Cryptomining & VMware ESXi 5.5 Update 2 Host Server

So, what has cryptomining got to do with VMware?

Late last year, when the bitcoin was around the US$600 mark, I embarked into cryptocurrency mining.  This was where I used my desktop together with some software like cgminer and began scrypt number crunching using my video card. During a couple of months of trial, I was mining Anoncoin, then moved on to Novacoin, and dabbled briefly on Peercoin which really didn’t work out. There was enough justification to go into this in a bigger way, i.e. 5 mining computers instead of one. I bought a few video cards, actually not a few, 3x Radeon 7950 cards, 7x Radeon 7850 cards, and a Radeon 7870 card. I even pressed into service my older Radeon 5850 card when gave up the ghost after its fan failed one day, but I replaced the fan and heatsink with an after-market cooler and kept it workng.  I played around with a lot of other cryptocoins – that is until the returns from mining would not cover the cost of our expensive electricity.  In addition the room was getting quite hot and having to have the aircon running during summer was just not acceptable. Okay – basically everything was shutdown in June this year, so now I have this hardware sitting around essentially doing nothing.

“Retask.it” – the mining hardware, of course. My VMware ESXi 4.0 Host Server was getting old, having run for several years and perhaps now was an opportunity to “Replace.it“. The current version of VMware ESXi is 5.5 Update 2. I put together some hardware to test this version – and had lots of issues installing it because some previously working hardware was no longer supported. There is another story there that I might tell another day. Anyway, after creating my customized installation cd that contains the Realtek 8168 network drivers, and updated adaptec array controller drivers – I was ready to install the production server.

The current configuration for my server contains the following parts:

Asrock 970 Extreme 4 AMD AM3+ motherboard with AMD Athlon X3 420e triple-core cpu and 8GB of ram.  The motherboard can handle up to 64GB of ram, so is sufficient for future expansion. There is no onboard video so I had to buy a single slot Gigabyte GV-R545 video card which houses a Radeon 5450 for $33. I don’t need a high performance card, just one that is a low power card.  The disk storage is an Adaptec 5805 Sata Array Controller (this was found for $300, normally $700+) – initially with 3x WD 3TB Nas Red drives, configured with Raid Level 1E.  I chose the Nas Red drives because they are designed for 24 hr operation – a little more expensive but hopefully are worth it, only time will tell.

My server needs multiple network cards, one for onboard management, one for internet connection, one for general network and one for backbone network.  Backbone is where I plan to have multiple host servers communicating – not implemented as yet.  The motherboard only has two PCI slots, so I could only install two network cards.  I have a couple of PCI-e networks cards on order – one of those will be for the backbone network.  I found from experience that having a few drives running 24 hours a day has a bit of heat, which requires a bit of cooling.  To that end, I have reused the Antec 1100 case to house all of these items.  This case is a very good for gaming and has lots of cooling, apparently better than the Corsair 500R that was also available.

One more thing is missing, the power supply – I have two FSP Aurum Pro 1000W power supplies left over – one of these was pressed into service and should easily handle another half a dozen drives for future expansion.  Almost forgot – add a cd/dvd-rom drive – I need one in order to install from my customized cd.  To save power, I can always disconnect it after installation – a good idea as this will be running 24×7, since one of the virtual machines is a firewall that protects my internal network from the world wide web.

Current capacity is 4.1TB of which I have used just 80GB, so still another 4000GB to go. If I add five more 3TB drives in Raid 5, this will give me 12TB additional capacity.  In comparison, the old ESXi 4.0 server had 5x 1TB drives in a Raid 1 and a Raid 5 configuration giving me a total of 3TB.  I didn’t know at the time that if I had upgraded the firmware on the Adaptec 5405 Sata Array Controller, I could have achieved this capacity with only 4 drives in Raid 5.  The older firmware at that time only allowed a maximum array size of 2TB to be created.  This was one of the benefits that came out of my testing of ESXi 5.5 – to work out what can be improved.

Anyway, still more work to do. Need to sort out all of the virtual machines, work out which to keep and migrate those to the new server. Better get on with it, I guess.

[PS]  There is a very good reason for including the updated adaptec array controller drivers in my customized installation cd.  During testing and installing adaptec monitoring software, I found that the included adaptec drivers that are bundled with VMware ESXi 5.5 Update 2 did not allow array monitoring, so I had to install some updated drivers from Adaptec.  After doing this, each time I rebooted the server, the datastore went missing.  The datastore houses all of the virtual machines – if this was missing, no virtual machines can run.  It turns out that upgrading the driver caused the VMware to think that the datastore is now a snapshot.  We cannot run from a snapshot (which is like a copy or an image), so the only thing that could be done to fix this permanently is to resignature the storage, but that meant I will need to relink every virtual machine (like 20 of them) – what a headache, so it is best not to upgrade the drivers unless absolutely necessary which means – use the right drivers from the start.
The technical document is here

http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1011387

Resurrect.IT – HP dc7700 Small Form Factor PC

Back in my lab – sounds great, doesn’t it – ok, it isn’t actually a laboratory, but it is my
working area where I work on computers.  I have worked in a real laboratory though – the Sydney County Council Measurements Laboratory, at the School of Electrical Engineering, UNSW.  We worked in the area of precise electrical measurements, was NATA registered and cooperated on many projects with CSIRO.  Ever actually measured a nanovolt?  I even designed and built a frequency divider with missing pulse detection for a rubidium frequency standard.  Another thing we did was to measure the capacitance of a semiconductor junction in a transistor – never before measured, only theoretical capacitance – and the measurement agreed quite well. This lab is long gone as is the Sydney County Council, but I digress.

Ok – the job at hand, to access stored browser passwords on the machine. One utility I have can access IE 7 & 8 protected storage passwords, but need to know the logon password, as this is used to encrypt the passwords.  The other utility can look for other browser passwords, but needs to run on that machine as that user.  Choices galore! In this case though, the Edwin user did not have a password, and I did not locate any internet explorer stored passwords. So now I need to “resurrect.it“.  Replacing the motherboard is possible, but after doing this, it might also be the power supply – or a combination. I could get a second hand machine on the internet – probably cost $100 plus, is it worth it?

I decided first to have a look at the power supply. After removing it from the machine, I opened it up and had a visual look – I found an electrolytic capacitor that was bulging, a sure sign of over-heating, but checking its ESR (Equivalent Series Resistance) – it wasn’t high. I didn’t have a direct replacement on hand, but I replaced it with one that was close enough but this did not fix the problem. I also removed the cmos battery – a CR2032, while I did more cleaning up. I removed the cpu cooler, took it apart to get rid of the accumulated dust, then checked the cpu, an Intel Core 2 Duo E6300. I cleaned the thermal compound off the cpu and heatsink, then applied Arctic Silver 5 and reinstalled the cpu and heatsink. After putting everything back in, applying power did not give the beeps and this time, the power stayed on. The machine still however failed to boot from the hard disk – but it was certainly working better than before. My diagnostic card that was plugged into the PCI slot did not indicate anything other than that the voltages were nominal.  This often means that the motherboard has failed – but it could be the bios has been corrupted. In any case, continuing on further is not economical – but I could do this later just as a learning process.

I could “resurrect.it” by virtualizing the machine.  Is this like virtual reality, you might be thinking – yes, very similar.  That is what I started doing yesterday. I have currently two VMware servers running, a new ESXi 5.5 server that I am configuring, and an old ESXi 4.0 server that runs my other virtual machines.  The first thing to try is whether or not the disk image that I collected previously would boot directly on VMware – it is unlikely, but is worth a try since it doesn’t take long, but I do need to bring the disk image into VMware first.

Step 1. Create a new virtual machine on esxhost2 (my new ESXi 5.5). The machine is called tdc7700 (for testing). The machine though is empty and I need to populate it.

Step 2. Boot the tdc7700 vm with a Ubuntu cd – once that is running, copy the disk image from the network onto the disk. I use dd again but before doing this I had to mount the network share using the mount -t cifs command.  Ok – done.

Step 3. Modify tdc7700 settings to make the disk independent, and not allow changes to be written.  This is to avoid having to copy the image again if something goes wrong.

Step 4. Power on tdc7700 – I saw the Windows XP screen come up briefly before a blue screen. Ok – thought it might happen.  This is probably because the original machine had a sata disk controller, and now tdc7700 has a scsi disk controller.  It cannot locate the original disk, hence the blue screen.

From here on, I will need to do a P2V conversion – a physical to virtual conversion, where in my case the physical is actually this tdc7700. There are a number of ways to do this, but the best way is to run the VMware offline converter which unfortunately is obsolete, but still usuable.  The converter is smart enough to make changes to windows hardware configurations in order to create a disk image that would run in a virtual environment.

Continuing, then…

Step 5. Boot the tdc7700 vm with a coldclone303 cd. Press the appropriate buttons to get to a target virtual machine called vdc7700 – it wouldn’t work to access esxhost2, strange. Everything is fine until it verifies the destination environment, then stops with an error. My esxhost2 is a newer VMware host and the datastore version is VMFS 5 – maybe this is causing a problem? So tried again, but now going to esxhost1 (my ESXi 4.0) server whose datastore is VMFS 3.  Success – I start the importing of the machine and it took about two hours to run.  Not bad with processing 80GB of disk across the network.

In the meantime, I was able to find a newer version of coldclone being 4.1.1, so when this finished, I tried again with coldclone411 but it still wouldn’t accept a destination of esxhost2, so decided to continue anyway, and create another virtual machine called vtdc7700 as it was time for bed. This time, I elected to minimize the size of the disks for the import process. once it had commenced, I went to bed.

This morning, I checked that the import had finished. Yes, it had – and now, will it work?

Step 6. Power on the new vdc7700. Yes, it boots up – at low resolution anyway. It tries to
update some drivers for changing hardware, that’s ok. I rebooted, then logged in and installed the VMware Tools – this ran, but then the machine hung – not so good, but not too bad.  Reset it and tried again, then successful. Changed the resolution to 1024×768 – much better. It has been resurrected!

Step 7. Run my utility to check for IE passwords – none, what? Check other browser passwords – none either.

I can access the machine in my virtual environment, but a lot of the programs are in Chinese, and I don’t read Chinese. It looks like 360Safe is installed. They are not using Internet Explorer so might be using the 360Safe browser to access email. Anyway, I asked them to come and see me and maybe they can show me how they access their email.

So in the meantime, a bright idea occurred to me – I could convert it so that they could run it on one of their other computers by using VMware Player. Since it is running now on my VMware host, it should be able to run on their computer. I will test this out and come back with a later update.

[PS]  Virtualization is quite good for running old machines where some application is still needed that cannot be moved to newer hardware.  Windows NT was always very picky when it comes to hardware and doesn’t cope well with newer hardware – so is very often moved to a virtual environment like VMware to keep it running. Windows XP on the other hand is much better at this and another alternative could be to put the hard disk into another computer and it will just update hardware drivers, maybe even have to activate again due to too many hardware changes at once – maybe I should do this and get rid of that old computer that is sitting around…