Hello,
Actually I’m running some 10 nodes with different public IP running on docker on same physical server each on different HDD of 14TB, since I have some nodes full and some other with 10-15% used I was thinking to move on big RAID5 configuration with 11 HDDs; using this configuration I can assign more space on node that need space and reduce the space allocation for the others.
What are the pro and cons ?
You’ll be adding redundancy the network doesn’t need, reduce the overall space you can be paid for, and make any IO issues worse (as RAID5 is one of the lowest-performance configs).
But yes you’ll gain some flexibility in how your nodes use space, and you’ll be much less likely to have a storage failure that would cause you to have to restart a node from scratch.
I think you’re better off running one-node-per-HDD until the HDD fails, and trying to get paid for every spare bit of space (instead of using some for resiliency).
1 Like
@Alexey what is the guidance on this? Ok or not ok? If it’s now OK I can fire up 17 VPN connection and host 25 nodes on one server today.
1 Like
It’s not ok, and you know this. This is evading the protection rule for nodes selection from each /24 to do not store the pieces of the same segment in the same physical location.
@brizio71 if your server would go down, the network would have a higher risk of lost or unrecoverable segment, and this could result in network shutdown, because the lost data = the lost customer = the lost trust = no payments = no payouts to anyone.
At best, the payouts will be cut even further, making the bypass uneconomical. Do you really want either of these options?
Regarding the topic, using a one disk per node is a best choice usually, but since you increase the risk of loss or made the segment inaccessible by bypassing a /24 limit, I would suggest to use a redundant setup for your hardware, including, but not limited to disks, the server itself, the redundant power supply or at least a managed UPS, the redundant internet connection, so using multiple ISPs, not VPNs, etc.
3 Likes
besides you shouldn’t be running one physical server with multiple IPs per the terms of service…
… also the maximum practical node size is around 18 or 20TB. If you start building a monster array it will be larger than what a single storj node can handle.
One node per drive gets you the best performance because the workload is so random I/O intensive.
1 Like
8 posts were split to a new topic: Using multiple IPs on the same physical service
A post was merged into an existing topic: Using multiple IPs on the same physical service