Unsolved

This post is more than 5 years old

3 Posts

32945

November 3rd, 2015 04:00

PowerEdge 2950 Revision 3

Hi,

I have a Power Edge 2950 server which I acquired a year ago for a home lab using VMware. It had 16GB RAM (using 8 x 2GB DIMMS). The system was rock solid with no errors reported.

Recently I decided to upgrade the memory from 16GB (using 8 x 2GB) to 64GB (using 8 x 8GB) and this caused issues with VMware both when booting from an existing installation on the HDD and via the CD-ROM media. Basically the system is going VERY slow when processing the "s.v000" file and the screen display refresh rate is also slow. Eventually after 40 minutes, all the VMware files are loaded but the machine crashes.

If I take out two banks of the RAM leaving 48GB in the machine, VMware will boot from the HDD and CD-ROM media and is orders of magnitudes faster processing the system files. The screen refresh rate is also noticeably faster.

I have also swapped the RAM modules around and have convinced myself that it is not the RAM. I don't think it is the slot either as the system was working with 16GB RAM before. I have tried this with VMware versions 4.0, 5.0, 5.5 and 6.x all with the same result. This seems to only affect VMware, as booting the server with Linux seems to work with all 64GB RAM!

The BIOS version is 2.2.6 and I have noticed that the latest BIOS is 2.7 - is it worth flashing? Could this fix the problem? Am I in any danger of bricking the server? 

I could just leave two sticks of RAM out of the server, but I would like to use all 64GB if possible

Many thanks in advance for the help

Mark.

Moderator

 • 

9.6K Posts

 • 

42.5K Points

November 3rd, 2015 09:00

Hi,

What version of Vmware are you using? It wouldn’t hurt to update the BIOS, but there is nothing in the release notes to indicate that it will resolve the issue.

3 Posts

November 3rd, 2015 10:00

Hi Josh, Thanks for the quick reply! I have tried numerous versions of the VMware bare metal hyper visor 4/5/5.5 and 6 all with the same result. Something I didn't mention last time was when the program crashes a CPU parity fault occurs on the front panel display. Doing a power cycle removes this issue and does not appear again unless booting VMware. Booting the same VMware image with 48gb ram (6 of 8 banks populated) it works without any errors or sluggishness. I did try Oracle VM hyper visor with the 64gb ram and all is good and no CPU parity error or sluggish display. It certainly seems related to the BIOS and/or due to the maximum amount of memory reached. How hard it is to flash the bios and how would you do it given there is no operating system installed? Is there a danger of bricking the server? Should all the auxiliary bioses also be updated ( eg raid/management)? Any further thoughts? Many thanks Mark.

11 Legend

 • 

16.3K Posts

November 3rd, 2015 11:00

The BIOS version is 2.2.6 and I have noticed that the latest BIOS is 2.7 - is it worth flashing? Could this fix the problem? Am I in any danger of bricking the server? 

You are in "virtually" no danger of bricking the server. This almost never happens. Can it? Yes, but it is extremely rare, and is probably more a result of user error (unplugging, cross-flashing, unreliable UPS, etc.) than anything. There are many undocumented changes in firmware updates, so it is completely possible that it will improve compatibility without stating specifically so in the release notes.

Honestly, I'm not sure it will fix your issue either. Is the RAM 100% compatible - does the RAM meet ALL the specs of certified RAM - speed, layout, timings, etc.? If not, it may work as long as it doesn't have to deal with too much of it. For example, there is one server (can't remember which one) that requires 800MHz or 1066MHz memory ... two sticks of 1333 will work, but 4 will not. Other machines have limits on ranks, speed, etc., so if it is not 100% compatible, it could cause odd behavior.

Moderator

 • 

9.6K Posts

 • 

42.5K Points

November 3rd, 2015 12:00

3 Posts

November 4th, 2015 16:00

Thanks everyone who has replied to my post. After more research the plot thickens.

When the server has the full 64gb ram, it appears that the bios is not reporting it all to the host operating system. At boot the server reports 63.54gb ram but the operating system is around 1gb. This could be the reason why VMware is choking when attempting to boot whereas a Linux Oracle VM installation didn't. This was confirmed when the oracle vm server have warnings that only 7mb was left of main memory after booting.

Take out the two 8gb sticks to give 48gb ram and all the memory is reported to the operating system. I have booted the diagnostic Live cd and confirmed that the system board is a PE2950 revision 3. I also went onto sell support entering the service tag and looking at the original configuration. Open manage does say that the max memory is 32gb with a memory array count of 1. However the server appears to work fine with 48gb ram albeit in a non optimal mode.

I have 2xE5440 CPUs installed and bios version 2.2.6. Open manage does recognise that my raid firmware is out of date but nothing reported about the later 2.7 bios.

I hope this helps

Mark

11 Legend

 • 

16.3K Posts

November 4th, 2015 20:00

It only gives warnings about out of date storage drivers/firmware. I would strongly recommend updating the system firmware (BIOS, ESM/BMC, etc.). Both the PowerEdge 2950 II and III support 64GB RAM; if OMSA gives you a max of 32GB, I would guess the BIOS needs to be updated (and best to update OMSA afterwards).

No Events found!

Top