Memory upgrade failed

[Steve Sokolowski]

The memory upgrade that was scheduled this afternoon failed. Chris determined that the motherboard is not compatible with 32GB modules.

He's going to need to build entire new systems to resolve this problem, which will require a week to get all the parts. In the meantime, Chris will restore all servers back to their previous configuration and get the system back online in 5 minutes.

[Murr808]

Is there no certified memory list for the motherboard?

[pfloyd1934]

Thank you for the fast response!

[mchat]

The memory upgrade that was scheduled this afternoon failed. Chris determined that the motherboard is now compatible with 32GB modules.

He's going to need to build entire new systems to resolve this problem, which will require a week to get all the parts. In the meantime, Chris will restore all servers back to their previous configuration and get the system back online in 5 minutes.

Thank you for the prompt update on this matter, its is MUCH appreciated. Please keep us informed on any time line changes, we continue to support ProHashing ! I've used this vendor for prior server memory purchases https://www.datamemorysystems.com/

[Marduke]

Steve, curious.....

Is the mining server just some type of suped up PC motherboard running 'NON-ECC' RAM or is this a true server type motherboard running standard server 'ECC' RAM.

If possible could you put up the specs or better yet a screenshot of the sticker on the RAM that is currently installed.

[Steve Sokolowski]

Chris did more investigation into this issue.

The current processor in this system is a seventh generation processor. It appears that Supermicro never bothered to consider the memory might ever get to 32GB per DIMM when this motherboard was released in 2014, or whenever it was released. There simply is a check that rejects all large memory modules.

We decided to upgrade to an entire new system. Chris will be purchasing a tenth-generation processor, in line with our policy of purchasing slightly-obsolete equipment. We do this because it costs up to $100,000 to buy a brand new server, but servers just 3 years old can often be had on eBay for $2,500. That much money will now get a computer with all server-grade components, redundant power supplies, 16 cores, a quarter terabyte of RAM, error correcting codes, and more.

We'll move about five years forward in technology, and be able to consolidate two four-core servers into one 16-core server with 256GM of RAM. It will use less power, so that the battery backup will give Chris more time to react in the case of a power failure. The advantage of this new server is that there will be no downtime. Since the mining servers do not store any data locally on their own disks, we can build the new server, make sure it works, and then disconnect everyone for five seconds to reset their miners.

We expect to have all the components in two weeks. In the meantime, four of the mining servers have not restarted in two days, so the new code has resolved the memory leak. Therefore, only half of customers are now significantly affected because the second server has less memory. Additionally, once we are able to finish the release of the latest version of the mining servers, restart times will be around 3 seconds, making any service interruptions minimal. By tomorrow, I don't think this issue will be bothering anyone much at all, even though the new server will not yet have arrived.

[UrbanMiner]

Thanks for the update and explanation.