New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Comments
Life is here and now — Dalai Lama
“Nothing is permanent. Everything is subject to change. Being is always becoming.”
Yeah. 3TB for $15/y is not very sustainable.
For your information, the average price for a Western Digital 3 TB hard drive in Romania ranges from 830 RON to 1,400 RON (approx. €170 to €280), depending on the specific model line.
Let say, you offer 20% existing-customer discount too:
Monthly billing: $5.59/month/TB
That would be 3x5.59= $16.77/month.
16.77x12= $201.24
$201.24 vs $15/y is not very sustainable for the future isn’t it?
Hmmm
hahahaha.
Best luck with your sales sir, thx u for 2y.
🍌
Happy we had you onboard.
And what exactly should I do with one 3 TB drive?? put it in a glass case?
, yeah, that would be the right logical choice 
Even an 8 TB drive is not particularly useful by itself. To build local storage properly, we need 24 or 26 drives, depending on the chassis we acquire.
Also, please Google why SATA drives are generally a poor choice for large arrays and why enterprise SAS drives are preferred. Then look up the price of an 18–24 TB enterprise SAS drive. I would rather not explain the same thing for the 100th time.
The bare minimum for a proper storage node should look something like this:
Now add:
Once you have calculated the acquisition cost, we can begin discussing operating expenses. So far, we still have no IP's, ASN, transit, router, switch, DACs/AOCs/transceivers, rack space, cooling, electricity, software licences, accounting, taxes, security setup, disaster-recovery procedures, spare parts, or staff — and one person cannot operate everything indefinitely.
Oh, and to build something remotely resilient, you need at least three nodes to get started, as if THE one you have decides to leave the chat, well, you are fqed
, need at least 1 on hot stanby to migrate / restore / or re-provision services the next day in the morning.
But yes, one retail 3 TB hard drive is cheaper.
, much cheaper 
Raid controller? Oh boy. Oh lord.
For local storage, in our opinion, only HW RAID. As you wish to have the lowest posible latency on storage and I will not trust a software setup to hold anything more then my porn collection on 24 drives, regardless of how many of you love ZFS ( we love ZFS also, just not at that scale )
For remote storage / SAN, that is something alse, but there Proxmox is out the window, and we would be entering the world of enterprise hardware and licensing.
Now in that case HA would be " Enabled by Default"
NetApp or Dell EMC would be the starting point, combined with VMware or Hyper-V and storage delivered over Fibre Channel, iSCSI, or potentially InfiniBand. Fibre Channel would personally be my preferred option.
However, that discussion is largely irrelevant here because the resulting service would most definitely not carry an LE* price tag, not by a mile miles.
Im not a fan of zfs, nor do I use it^^ but you really have to get rid of it.
Get rid of what? ZFS, we don't use it in PROD, we have 2 servers with 4 bays with ZFS for scanned documents and ISO images and company archive, other then that, fuq no.
As for SAN, yah, we do have a NetApp, older model + 2 shelvs and Brocade FC switches, sucks power like a 5.7L V8, but man that think rides like a rocket, but not for LE* products.
Also, replacement drives from CISCO cost as much as a Liver Transfer. not to mention the cost of "features" enabled in that controller.
Either way, reply me on the first question, I might have not got it..
Hi,
if you ride on the lonesome warrior hostnode with local storage, then hardware raid is the option that will give you most performance.
Its worth to think about using it with a filesystem that does checksumming.
In basically all enterprise environments you will not find hardware raid controllers, simply because they are a bottle neck and its not clear if the data was really really really saved on the disk then then ack'ed or if its somewhere in DRAM about to be saved.
Thats why basically all storage (cluster/scaling) software products recommend NOT to use hardware raid controllers. Some even refuse to work with them ( MS S2D for example ).
But if you do not build scaling storage then raid is the way that will cost you the lowest amount of performance.
So for cost effective stuff, thats the legit way to go.
What's especially worth noting and I want to add here is that you need to get a exact replacement and pray to the gods it reads the disks fine(let alone that you have a single point of failure). Money well spend when software raid does it better.
not necessarily and it depends on a lot more factors.
It is not a simple yes and no situation.
As a general rule:
Low cost services = Local Raid, and in that case HW RAID card in our view
High end services that need HA = cluster storage / SAN
but it is not necessarily related to low/mid/high type of service, it depends on the use case and necessities of the customer, or of that service in particular.
If you wish I would dive into specifics, but we need a bit of a context here.
A Ferrari might be a fast car, but on County side roads full of potholes a Lada Niva will get you farther and faster, as an analogy.
If you wish to extend on this, just say so.
@Mainfrezzer
Did EXT4 journaling save you on your issue?
Was that a 2 drive raid1? or a 3 drive raid 5?
Were all of them SATA drives?
Did you use the onboard AHCI controller or a HBA card?
Some of us over here living dangerously with unraid and some random hba card flashed in IT mode. Tho I at least finally moved to dual parity....
If you're refering to my post with the vps search, yes ext4 fixed the issue, although that's not the usual filesystem I use, it just on that vps and vdisk.
I've no clue what the provider runs in the backend but the issue was ultimately that I got migrated to a wrong node which caused the issue, since then it's been fixed.
For my proper nodes I use btrfs(certainly not raid5 or 6 lol), they use hbas
Oh, the good old days flashing LSI 2008 cards to IT mode......... ah, good old days.
I would indulge you to go pure raid 10 and just drop the cache ( for safety reasons ) unless you have 5KW of backup batteries and the box uses 80W in full load
.
Stay safe !
OK, i saw a DM0 error there, hence I asked, I resumed you were running some kind of local sw raid.
I can confirm you some weird issues in Prox 9.1 and 9.2 on live migrations with local storage, we saw some a few moths ago, since then we switched migrations to offline for the moment, yet with kernel 7.X we could not replicate the issue.
I swear, each CVE update fixes one issue and opens 3 more in 2 weeks.
Glad to hear you managed to fix/solve your issue.
The more I read, The more poor I feel. @host_c a single 36 drive cost as much as a car
Seems dangerously close to running out of space 😂 clearly time to expand, drives are cheap * run *
Just what until you see what an enterprise core router costs.
Yeah, that's another whole crazy that makes your jaw drop... Ah, "carrier grade"
host_c mentioned Brocade earlier. Some of their stuff can be dirt cheap on used markets if you only need 10-40Gbps ports, but even 100GbE can be good value. I've used a few of their older switches
(some models are still getting updates)
Don't forget the licensing.
And the support.
Great point. A lot of them are honor-based but expensive
Now could be the best time to sell the equipment, ride the bubble before it pops. One can simply restart the business afterwards and buy equipment for cheap. However, nobody knows the future; we just speculate based on a loop history.
Key point to think about before longing/shorting anything in life.