Hello,
>Actually, VMware is not less tolerant of Memory errors, it is just that it USES all memory in the system unlike other operating systems. So if you had say 8GBs of >memory in a Windows box, the chances of you touching all memory in the system is pretty slim unless you are running something very large. So since ESX touches >and uses all the memory depending on load, what you see is different. A few things with the DLxx5 series. Make sure your CPU/Memory boards are balanced. I.e. do >not have 32GBs on one board and 16 on the other. VMware definitely does not like that!
Not my idea this is coming from HP's service, even their escalation engineer was saying this,
>Also you could have bad memory, it happens from time to time. WIth the density of memory it was discovered that Shielded memory tends to work better than non->shielded. Also make sure the Memory is HP Branded.
The memory was factory installed by HP. Even though it may be HP branded, the memory may be from any number of OEMS such as Kingston, Micron, Samsung, Hi-Link,etc. Their techs are saying that they are finding they are having better luck when replacing DIMMS if they replace them in pairs and use memory from the same OEM, At one point they were suggesting that the problem was with memory from one particular OEM but then later backed off that claim.
This is actually kind of humurous. I had a big dust up with HP before about obtaining additional memory for 2 DL 385s. I ordered memory from HP to upgrade these boxes from 8 to 16 GB. After receiving the order the reseller contacted me to tell me that HP couldn't provide these parts anymore and that they would have to obtain the memory second source from Kingston. I objected strenously to this to no avail. I almost cancelled the order for the 585 G5s based on this, Maybe I should have? I ended up ordering the upgrade memory myself from Crucial. These two servers have had no new issues after adding this memory and installing 3.5 update 1. I did have to have 2 of the original HP branded dimms repalced though. :smileydevil: (There had been internal health warnings on this server before the memory was added.)
>It tends to work much better than non-HP memory.
>I would definitely make sure your memory is shielded, but other than that, it is possible you got several bad sticks of memory.
>As for testing, you want to run memtest86 and not the HP diags. Run them for at least 24-48 hours.
Their escalation engineer ran their own proprietary meatgrinder utility on one of the servers all weekedn after replacing all the memory on one of the 585s. They only ran it for a couple of hours on the other and we ended up having another failure after booting it back up in ESX within 24 hours.
>And do the same for the HP Diags before you even place ESX on
>the system.
I ran HP diags on a DL 385 G5 all weekend beofr putting ESX on it. A dimm failure occurred after 5 days of running ESX 3.5. They replaced the DIMM and it's partner and it has been OK since.
>I saw quite a few issues with memory when working at HP, and most of the time it was >shielded vs. non-shielded. At least in really dense configurations.
I'll look into this but I didn't see any option to choose shielded memory when configuring these servers.
