Help picking hardware for Trunas VM Storage server

[jgreco]

I can't speak to any historical issues, but I'd hope that most of our contributions these days are positive.

I did say "during the early years". There was this thing where I was trying to advocate against AMD APU based mini-ITX 2GB-4GB RAM systems especially where people were trying to run ZFS (UFS was still available at the time) because there was circumstantial evidence that such systems would kernel panic in some circumstances and leave an unusable pool; the pushback I received from iXsystems was pretty bad but I *was* allowed to edit the user manual to establish the 8GB minimum memory requirement. There were other more problematic pushbacks such as the 3Ware and Areca RAID controller issues where I appreciated the idea that "this hardware SHOULD work" but junky RAID controllers quite simply had a different set of design goals. It is easier to explain this to users by suggesting that they should design systems that more closely resemble what iXsystems would sell them, rather than hoping that their janky old 4GB Opteron with a 3Ware RAID controller was going to make a good FreeNAS, even if it was hot hardware in 2005.

There did come a point at which Dru Lavigne apparently made a compelling argument to the dev team that we here in the forums were doing the heavy lifting of tech support for the free product. Things got quite a bit better. That and Cyberjock being hired, since many people viewed him as an abrasive presence, and hiring him gave him something better to do. Probably went a long way towards smoothing things out.

The current status quo seems to be generally reasonable -- it's nice having old forum members like @anodos hired by iX taking care of some of the more complicated issues, especially ones like Samba that tend to be moving targets. You're very active here as well, I think primarily handling SCALE stuff, which is helpful given that some of us aren't Linux-heads. Moving forward, though, I think you might need to encourage more iX participation here. The volumes seem like they have shot way up in the last two years, and the number of SCALE experts hasn't SCALE'd accordingly.

Just random stuff to think about.

 

[morganL]

The current status quo seems to be generally reasonable -- it's nice having old forum members like @anodos hired by iX taking care of some of the more complicated issues, especially ones like Samba that tend to be moving targets. You're very active here as well, I think primarily handling SCALE stuff, which is helpful given that some of us aren't Linux-heads. Moving forward, though, I think you might need to encourage more iX participation here. The volumes seem like they have shot way up in the last two years, and the number of SCALE experts hasn't SCALE'd accordingly.

Just random stuff to think about.

I'm optimistic that SCALE experts will grow with the user base
AND
SCALE problems will reduce with code/documentation maturity. This week will be a major step.

 

[Adia113]

I got curious about chatgpt right around the time I started building my nas. I am not saying it is always accurate but it does help with painting the broader picture if you know how to ask it the right questions. I checked everything it said to me using other sources and only asked it things that I looked up a little before hand. I will say I notice it gets the location of where things are in Truenas incorrect, so I do not ask it to show me how to do things in Web UI. I have not tried the new paid version so it could be more accurate now.

Research takes up a lot of time when trying to learn on your own without taking courses or having someone show you. Forums like this are a great resource but they take time and people dont read the posts they just ask their question. Then you all have to answer the same question 100 times. I think chatgpt, if it becomes more accurate will be good for asking all those basic questions and then coming to a place like this to discuss more complex ones or to verify answers.

In my first post I created many different trees for my dataset and asked chatgpt which ones it thought would be good. It had some good suggestions but also some bad ones that didnt make sense. I asked it how I would add jails to the layout and the way I asked it made it answer me incorrectly. It wasnt incorrect for how it proposed I setup the dataset, it was incorrect for not telling me I was wrong for wanting to add it. I then brought that layout here and Samuel Tai edited it into something that would work better. chatgpt was right in how it presented the tree but it added unnecessary stuff because I added unnecessary stuff. That is what makes chatgpt crappy sometimes. I am curious to see what happens with it in the future.