Hello there i post this topic about gluster fs hosts for storj storage node.
In my case i have 2 hosts,
I do recommend to add another host so 3 hosts
I have the Docker container running on cent os 7,
So this solution for redundancy (experimental) is to replicate data in real time across hosts with network gluster fs.
If a host is lost data is already current (i hope) in the other hosts and i have to relaunch Docker container manually on other hosts.
I hope this solution will work wait and see.
I use a gluster fs volume as data path
Replication is up with gluster un realtime !
Regards.
Plus : please add Docker swarm capability (docker cluster)
@Alexey, will something like gluster let me mount my storage from many nodes to a single node where I could run a storagenode? I’m just wondering if I could like proxy all my storage from 8 different computers to 1 computer so I have 1 node running storj, but that node is using the storage across all 8 nodes? Also looked at Ceph… do you have any suggestions for my use case?
I can only suggest to make a 8 identities and sign them and run 8 nodes.
But I would recommend to start them one by one - when the first is filled up, then you bring up the second and so on. In this case it will not take forever to get vetted.
However, you will not receive more traffic than only one node anyway. The only plus is protect the risk to lose everything with a storage failure.
The storagenode itself is a node for the global distributed file system.
As far as I know, the local/small distributed filesystems such as GlusterFS/Ceph/etc uses mostly a replication to protect the data, so, the capacity will be lower than 8 separate disk/nodes. Also, any network storage will be noticeable slower than a local storage. In case of storagenode it will fail too many requests and will have a less data and traffic.
But if you want to experiment with GlusterFS/Ceph/etc - go ahead. It’s up on you.