MG2R

You’re good 

Data criticality means it needs stringent backups. RAID is intended for uptime/availability assurance, not loss prevention. 
 
is there a reason the accounting dept has more stringent uptime requirements vs marketing? Both need to be able to work. 
 
More stringent backups for accounting, that I get though. 

Although I could not figure out what your exact requisitions were, I would attempt to make some recommendations:
You may wish to make a NAS that was more powerful & stable. That's a good start, and given that the FX-8320e performed merely identical to the last-generation Pentium G6405 (only having 2 cores!), an upgrade to a more recent 4-core system could be considered. Take the i3-10100F & two 4GB memory sticks out of System 2, then pair it with a dirty cheap H410/H510 board. After that, the Core i5-11400F processor could be considered for System 2, since that your use cases would seem to suffice with 6 cores, not 8. The GT 730 GPU would seem to be useless for such a server. Use it only when installing a system, then pull it off. If these 2TB drives were still healthy, they could be reused by building a RAID array, such as RAID 6 or, less stable, RAID 5. However, this would end up having only 4TB of available storage, and new drives with large capacities could be considered as needed. Upgrade both System 1 & System 2 with more RAM & more SSD storage, since they are nowadays pretty affordable. You may wish to take a try on machine learning at home. This would require a quite more powerful GPU with a huge amount of VRAM, and there were 2 options: a) a regular GPU like RTX 2060 or 3060 12GB, installed onto either System 1 or 2, since they were capable of gaming as well; b) a used Tesla card like P40, installed onto the NAS, as they performed well in computing, and generally had massive amounts of onboard RAM. Hope these would help you make decisions more wisely.😉

When I saw the short circuit on this case, I knew I had to have it. 
 
I’m a stickler for water cooling so I did my best to jam a 3070Ti, a 5600X, a thick 80 rad, a 140 rad, and a slim 240 rad in there. All brown Noctua, as it should be. 
 
result came out ok but it did cost me literal blood and tears. The density of this machine is immense. It weighs so damn much for its size. 
 


this build was done in two steps, as the right tubing did not arrive and it was hard to source where I’m at. I built the whole machine with clear tubing at first. I did not take pictures of the initial build apparently. So here’s the rebuild with white tubing
 
original build 


first taking it all Apart. The joys of SFF. Here you can see the slim 240 in the bottom, there’s slim fans underneath. It clears the gpu block
juuuust right. 

 
added the cpu block again. Had to turn this one around as I apparently swapped inlet and outlet on the first go

 
 
I cut my finger with Jerry’s knife from lttstore while cutting tubing. It sprayed blood over my entire kitchen and I needed stitches. Jerry’s knife is damn sharp. 0/10 would not recommend. Part of my finger still does not register touch , nicked a nerve. 

worth it though. She’s b-e-a-uuuutiful

Came here to post about Synology. I have been beaten to the punch. Whatever you do. Don’t build it yourself. Get a solution from a reliable vendor. 
 
Although, this kinda sounds like an XY-problem. The marketing department needs a NAS? Why? What are there actual requirements? What made them come to the conclusion “we need a NAS”. Is that conclusion correct? 

The CPU thermals are great, 40C to 55C under load. But i did have to add extra fans to the RAM to keep all of it cool. Especially after spinning up some VMs on this box and RAM utilization goes up.
I ended up using 2 60mm noctuas which cover the RAM completely and i added some zip-ties to keep them aligned within the free space between the RAM sticks. 
 


 
I will probably have something 3D printed though to better funnel the air through the DIMMs. maybe just a shroud to properly funnel the air from the radiator over the RAM aswell?
I pay about ~25c a kWh, meaning my homelab is currently drawing about 200 euro a month worth of electricity. Quite a bit, but to me the hobby is worth it. My SO also isn't complaining because i'm paying the powerbills anyway.
 
As for the gaming rig, thanks! i built it after my previous pc burned down. That's quite a story in of itself, you can check out the Phoenix link in my signature for that buildlog, eventough the machine has seen some changes since then.
 

When I saw the short circuit on this case, I knew I had to have it. 
 
I’m a stickler for water cooling so I did my best to jam a 3070Ti, a 5600X, a thick 80 rad, a 140 rad, and a slim 240 rad in there. All brown Noctua, as it should be. 
 
result came out ok but it did cost me literal blood and tears. The density of this machine is immense. It weighs so damn much for its size. 
 


this build was done in two steps, as the right tubing did not arrive and it was hard to source where I’m at. I built the whole machine with clear tubing at first. I did not take pictures of the initial build apparently. So here’s the rebuild with white tubing
 
original build 


first taking it all Apart. The joys of SFF. Here you can see the slim 240 in the bottom, there’s slim fans underneath. It clears the gpu block
juuuust right. 

 
added the cpu block again. Had to turn this one around as I apparently swapped inlet and outlet on the first go

 
 
I cut my finger with Jerry’s knife from lttstore while cutting tubing. It sprayed blood over my entire kitchen and I needed stitches. Jerry’s knife is damn sharp. 0/10 would not recommend. Part of my finger still does not register touch , nicked a nerve. 

worth it though. She’s b-e-a-uuuutiful

I just had a look, The pictures are from November last year. There’s still a spot in my finger I can’t feel and pressing on the place I was cut gives me that tingling feeling you get when your leg wakes up after sitting too long
 
good thing is the dull spot is on the outside in a place where you don’t really need sensation so my day-to-day is not affected at all. The sound itself healed up rather nice as well 🙂
 
I can’t believe it’s been that long already though, wow.
 
It’s freaking tight but it works. Biggest issue is that I had to use a DDC pump instead of a D5. It’s rather noisy and I don’t have the space to spring mount it securely.
 
Overall performance is reasonable. I have it dialed in for silence so my temps are rather high, but the system is almost inaudible during regular operation. 

the performance is actually more than fine.

Better airflow and static pressure than the Noctua A12-25, but they're louder.   So, tradeoffs.  

I do agree it looks badass having counter-rotating fans in a row of similar fans.

Y’all are overthinking this. OP is not talking about reversing fan flow. They just want the blade to point and spin in the other direction. If you reverse both the blade sweep and the motor spin you still move air in the same direction as a conventional fan. 
 
to make it more clear: if you look at a fan from the exhaust side, almost all will turn clockwise and have their blades swept clockwise. OP wants a fan that spins counter clockwise and has its blades swept counter clockwise. 
 
OP, the reason is cost. Motors spin clockwise by convention. Procuring motors spinning counterclockwise has added cost. Similarly, keeping two SKUs is more expensive than one. 
 
edit: as I’ve posted this I was re-reading. There really should not be a cost difference as dc motors can be reversed by changing the polarity. Bldc motors too by changing the order of the coil activation. I still think there’ll be a cost difference as more skus==more expensive, both for motors and for blades, but it should be minimal.
 
edit edit: I mixed up clockwise and counterclockwise above. Fixed. I should go to bed.

I would say Ceph is probably a little too complicated solution for this, Ceph-Ansible does make the deployment a lot easier but if you need NFS or SMB protocol you need to install extra gateway software etc, quite a lot of layers. Also it's not considered all that "safe" to do singe server Ceph either, you can change the deployment option to allow it since by default that is not supported but it's for lab testing only really.
 
Single server GlusterFS would probably be middle ground between LVM RAID and Ceph, less complex and better native NFS/SMB capabilities. Still not recommended for single server either.

Everything was set up correctly. It’s just a symptom of shared IP addresses with mail servers. Not Mailgun’s fault really. They had anti-spam detection from what I remember but nothing is 100% watertight. 
 
as soon as you pay for a dedicated ip, those issues go away. Again, to their credit, any time we noticed such an issue they had us switched to a different IP immediately.

Used them at a previous employer. At that point they were our cheapest option and it showed. We had it happen multiple times that the sending IP we were assigned was shared with someone who sent spam and thus our mails were getting blocked. 
 
Of course, if we would’ve paid for a private IP we wouldn’t have had those problems. To their credit, they were responsive on their support and we were never annoyed enough to switch providers. 

Used them at a previous employer. At that point they were our cheapest option and it showed. We had it happen multiple times that the sending IP we were assigned was shared with someone who sent spam and thus our mails were getting blocked. 
 
Of course, if we would’ve paid for a private IP we wouldn’t have had those problems. To their credit, they were responsive on their support and we were never annoyed enough to switch providers. 

Used them at a previous employer. At that point they were our cheapest option and it showed. We had it happen multiple times that the sending IP we were assigned was shared with someone who sent spam and thus our mails were getting blocked. 
 
Of course, if we would’ve paid for a private IP we wouldn’t have had those problems. To their credit, they were responsive on their support and we were never annoyed enough to switch providers. 

Used them at a previous employer. At that point they were our cheapest option and it showed. We had it happen multiple times that the sending IP we were assigned was shared with someone who sent spam and thus our mails were getting blocked. 
 
Of course, if we would’ve paid for a private IP we wouldn’t have had those problems. To their credit, they were responsive on their support and we were never annoyed enough to switch providers. 

Thank you so much everyone. I think I have everything I need now for any future conversations with Protonmail.
 
Going to lock the thread now.
 
Greatly appreciate the help again, and hope everyone is having a good day.

Thanks guys this was very helpful. I've managed to set it all up using dataset quotas, replication tasks and NextCloud to achieve everything I set out to do. I appreciate the help and pointing me in the right direction

OP, keep in mind the bathtub curve. Hardware will usually either fail when it is almost new or when it is fairly old, with little failures in between. https://en.m.wikipedia.org/wiki/Bathtub_curve
 
To me, it sounds like you had a bad batch of drives. And it sounds like while working in your system you disturbed a drive that now has performed reallocation.
 
@LIGISTX is right here. Burning in a drive is definitely good practice, especially if you buy multiple drives at once. If they’re from the same batch and there was a manufacturing issue with that batch, you are jeopardising your data by putting it on a pool with a high chance of simultaneous failure.
 
About cascading drive failures: this is anecdotal but one of my first gaming rigs back in the day just ate hard drives. We’re talking every few months, it’d have a drive die. Back then I never figured out why and I ended up getting rid of the system. 
 
my running theory now is bad power. Others have commented on your PSU but the sticker tells you only how many Watts it’ll be able to deliver. It tells you nothing about the quality of that power. Too much ripple in the voltage can be devastating for  sensitive electronics like your hard drives. The unfortunate part is that there’s no easy way to test this without an oscilloscope. 

OP, keep in mind the bathtub curve. Hardware will usually either fail when it is almost new or when it is fairly old, with little failures in between. https://en.m.wikipedia.org/wiki/Bathtub_curve
 
To me, it sounds like you had a bad batch of drives. And it sounds like while working in your system you disturbed a drive that now has performed reallocation.
 
@LIGISTX is right here. Burning in a drive is definitely good practice, especially if you buy multiple drives at once. If they’re from the same batch and there was a manufacturing issue with that batch, you are jeopardising your data by putting it on a pool with a high chance of simultaneous failure.
 
About cascading drive failures: this is anecdotal but one of my first gaming rigs back in the day just ate hard drives. We’re talking every few months, it’d have a drive die. Back then I never figured out why and I ended up getting rid of the system. 
 
my running theory now is bad power. Others have commented on your PSU but the sticker tells you only how many Watts it’ll be able to deliver. It tells you nothing about the quality of that power. Too much ripple in the voltage can be devastating for  sensitive electronics like your hard drives. The unfortunate part is that there’s no easy way to test this without an oscilloscope. 

Like the multitude of other threads: run TrueNAS for a point-and-click, easy-to-use experience. Install your favourite flavour of Linux then run either Docker Compose or K3S for your containerised software if you want the CLI experience.
 
Whatever route you take, make backups first, dink around later. I like Borg Backup but there’s so many other ways to do it. Seriously, back up your stuff somewhere _before_ playing around with your NAS. 

OP, keep in mind the bathtub curve. Hardware will usually either fail when it is almost new or when it is fairly old, with little failures in between. https://en.m.wikipedia.org/wiki/Bathtub_curve
 
To me, it sounds like you had a bad batch of drives. And it sounds like while working in your system you disturbed a drive that now has performed reallocation.
 
@LIGISTX is right here. Burning in a drive is definitely good practice, especially if you buy multiple drives at once. If they’re from the same batch and there was a manufacturing issue with that batch, you are jeopardising your data by putting it on a pool with a high chance of simultaneous failure.
 
About cascading drive failures: this is anecdotal but one of my first gaming rigs back in the day just ate hard drives. We’re talking every few months, it’d have a drive die. Back then I never figured out why and I ended up getting rid of the system. 
 
my running theory now is bad power. Others have commented on your PSU but the sticker tells you only how many Watts it’ll be able to deliver. It tells you nothing about the quality of that power. Too much ripple in the voltage can be devastating for  sensitive electronics like your hard drives. The unfortunate part is that there’s no easy way to test this without an oscilloscope. 

Ok, so interesting. OpenZFS (what TrueNAS uses) is different in this regard than oracle ZFS. 
 

so it indeed is not as flexible as I remembered. I think I had been reading the oracle ZFS manual, which does not stipulate this problem. 

I guess this is the difference. I have backup on for everything, and wouldn’t want to lose that data. Backing it up any way besides via iCloud is just not feasible. I want my phone to be fully restorable, exactly how it was, in the event I need to go down that path. Thus my argument of replacing iCloud being impossible still holds true - any other option comes with sacrifice. If those sacrifices are acceptable to you, then that’s perfectly fine.