I also have a stack of sm863s, 20+ between my TrueNAS and proxmox nodes. I have some with 1700 write cycles on them. Not worried.
How many full write cycles are you seeing your failures at?
start/run gpedit.msc
Local computer policy > Computer configuration > Windows settings > Right click Policy-based QoS > Advanced QoS Settings > Inbound TCP Traffic > Specify inbound TCP throughput level > Level 3
And then adjust all of your 10gig NIC hardware options for max perf on both...
I had a 60mph storm come through the other night so all my hardware is still powered down... but there is a flag you can throw at fio-status and you can specify a variable to display a value for. I think you can specify optrom stuff and get info back from that. I think it's -k or -K or -F for...
So I tried a few different things for 1U with my 1541 just using some card stock and the noctuas. Having two pull air through the heatsink with a cardstock shroud still brought the CPU up to 90C under heavy loads.
3 fans and a cardstock duct finally had enough airflow to cap loads out at 80C.
That's like how iCUE kept causing pubg lockups since they used battleye as their anti-cheat. As soon as battleye would randomly scan iCUE during gameplay, everything hard locked =P Took me about a year to figure it out =P
Why not just run a narrow ILM Noctua cooler like this?
https://www.amazon.com/gp/product/B00E1JGFA0/
If your case is wide enough, you can run one of these as well: https://www.amazon.com/gp/product/B00DWFQ42I/
I use these on all my X9 and X10 supermicro boards without any issues.
If you just want more sata, get one of these $50 LSI clones off amazon and have a field day. I have a few of these in my lab the last 10 years and they are quite popular in the FreeNAS/UnRaid/Proxmox community.
https://www.amazon.com/gp/product/B0C2YMJHY3
I avoid raid in this day and age of...
Just bumping this up since I have VSL 4.x working in the latest proxmox 8.1 with no issues.
I didn't take exact notes, but reconstructing some of the fiddling in my bash history, this was the rough proces:
### Fusion-io install
apt-get install git make gcc dkms fakeroot build-essential...
3.1.1 for RHEL 6: https://support.hpe.com/connect/s/softwaredetails?language=en_US&softwareId=MTX_cdc387e878e54821a243515856
3.1.1 for windows: https://support.hpe.com/connect/s/softwaredetails?language=en_US&softwareId=MTX_8701f69a4971442abe3896de2a
Should be easy enough to get a centos 6 ISO...
Write amplication depending on workload is definitely a factor. Some workloads are such that writes can result in 4x the actual committed blocks on the underlying storage, so you'd go through that wear life 4x faster than expected
Well, the old Intel S3500 1.6TB sata drives will do 8PBW... All of those older MLC 1.6TB Intels will do that. I think the 1.92s are on the newer 3D NAND and are TLC... they should probably do about the same.
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.