OptiBiz1
Active Member
I am looking into our future server config need and I am currently looking at dual servers with a load balancer (it's actually primarily the failover that is of interest), but disk is a big question mark.
In our architecture we have one "control hub" that controls the network of instances and provides extensive (!) services to resellers of the hosted services + acts as a node that provides a query interface for external systems (which we will also provide). This "control hub" server needs to be a pretty badass server.
Each instance of the services as such (based on MW + our 53 add-ons + infrastructural components) is probably subject for redundancy as well.
Perhaps this config:
Xeon D 1540 (or possibly D 1520)
128 GB RAM (or possibly 64 GB RAM - any ideas on this one?)
2 * 2 TB RAID1 + 2 * 480 GB SSD RAID1
(I am trying to avoid separate storage. Yes, it would make the architecture more redundant and "clean", but it would also bring up the cost considerably)
Delivery servers will be some of the leading services out there, that is, we will not host any PMTAs etc ourselves.
The idea instead is that the SSDs will have the MySQL database on them while the regular disks are just sitting there with the scripts and components. MySQL in this case would be the replicated with the other server in the pair in case any of the two nodes in the server pair would go down.
I am not sure if we will need to replicate the regular disks. Do we?
1) Does any of you run your servers with a config similar to this? Any experiences you would be willing to share?
2) Do Mailwizz store things on disk on the fly which is not stored in the database?
3) If so, what would be a good way to keep the disk pairs up to date? Someone mentioned NFS. Is that fast enough? Good for the purpose? Any other ideas?
I know there are plenty of awesome people here on these forums that I am sure have interesting experiences and I'd also like to get some input and ideas from @twisted1919 if possible.
Thank you!
In our architecture we have one "control hub" that controls the network of instances and provides extensive (!) services to resellers of the hosted services + acts as a node that provides a query interface for external systems (which we will also provide). This "control hub" server needs to be a pretty badass server.
Each instance of the services as such (based on MW + our 53 add-ons + infrastructural components) is probably subject for redundancy as well.
Perhaps this config:
Xeon D 1540 (or possibly D 1520)
128 GB RAM (or possibly 64 GB RAM - any ideas on this one?)
2 * 2 TB RAID1 + 2 * 480 GB SSD RAID1
(I am trying to avoid separate storage. Yes, it would make the architecture more redundant and "clean", but it would also bring up the cost considerably)
Delivery servers will be some of the leading services out there, that is, we will not host any PMTAs etc ourselves.
The idea instead is that the SSDs will have the MySQL database on them while the regular disks are just sitting there with the scripts and components. MySQL in this case would be the replicated with the other server in the pair in case any of the two nodes in the server pair would go down.
I am not sure if we will need to replicate the regular disks. Do we?
1) Does any of you run your servers with a config similar to this? Any experiences you would be willing to share?
2) Do Mailwizz store things on disk on the fly which is not stored in the database?
3) If so, what would be a good way to keep the disk pairs up to date? Someone mentioned NFS. Is that fast enough? Good for the purpose? Any other ideas?
I know there are plenty of awesome people here on these forums that I am sure have interesting experiences and I'd also like to get some input and ideas from @twisted1919 if possible.
Thank you!
Last edited: