Hi,
In the company we have decided to migrate our applications from a public cloud provider to our own solution based on proxmox ha cluster. Since this is the first such installation I have a great request for an opinion on the approach we have chosen for this.
General assumptions:
- 3 identical servers (cpu, ram, storage) put together in ha cluster
- servers hidden behind 2 fortigate 60f firewalls
Hardware we plan to use i 3 identical serwers:
DELL R730xd with 2x rj45 1G, 2x rj45 10G
2x Intel Xeon processor E5-2690 V4 SR2N2 2.6-3.5 GHz
4x 64GB RAM DDR4
2x 400GB SSD
2x 12TB SAS
1x 4TB SSD
and 2 Fortigate 60F
The current plan is to:
1. connect the servers in a mesh cluster using 10G links, and 1G links with fortigate set to active-pasive mode
2. put the system on 400GB SSD drives tied together in RAID1
3. set up 2 ceph pools. The first pool (2x12TB HDD SAS) is to be dedicated for functions that require a lot of space, but are rarely used (low IOPS). The second (4TB SSD) dedicated for VMs operating system and applications with a high number of IOPS
My question to you is:
1. Will this solution generally work well as a production environment. Our VMs will host several web platforms, and databases. The whole thing is rather small, the growth in terms of data is about 4TB per year. Data transfer to and from platforms is probably around 500GB per month?
3. Is mesh cluster a good idea? Certainly beneficial due to the lack of need to buy expensive switches.
4. Will 10G connections on rj45 for ceph needs be sufficient, or is it better to think of faster connections now?
5. Whether such a solution can be reasonably scaled in the future. In terms of RAM or disks I know that there should be no problems, but what if you need to supply another server (or servers to keep an odd number of them). Would a mesh cluster still be a good solution in such a case?
Thanks for any feedback
Bartek
In the company we have decided to migrate our applications from a public cloud provider to our own solution based on proxmox ha cluster. Since this is the first such installation I have a great request for an opinion on the approach we have chosen for this.
General assumptions:
- 3 identical servers (cpu, ram, storage) put together in ha cluster
- servers hidden behind 2 fortigate 60f firewalls
Hardware we plan to use i 3 identical serwers:
DELL R730xd with 2x rj45 1G, 2x rj45 10G
2x Intel Xeon processor E5-2690 V4 SR2N2 2.6-3.5 GHz
4x 64GB RAM DDR4
2x 400GB SSD
2x 12TB SAS
1x 4TB SSD
and 2 Fortigate 60F
The current plan is to:
1. connect the servers in a mesh cluster using 10G links, and 1G links with fortigate set to active-pasive mode
2. put the system on 400GB SSD drives tied together in RAID1
3. set up 2 ceph pools. The first pool (2x12TB HDD SAS) is to be dedicated for functions that require a lot of space, but are rarely used (low IOPS). The second (4TB SSD) dedicated for VMs operating system and applications with a high number of IOPS
My question to you is:
1. Will this solution generally work well as a production environment. Our VMs will host several web platforms, and databases. The whole thing is rather small, the growth in terms of data is about 4TB per year. Data transfer to and from platforms is probably around 500GB per month?
3. Is mesh cluster a good idea? Certainly beneficial due to the lack of need to buy expensive switches.
4. Will 10G connections on rj45 for ceph needs be sufficient, or is it better to think of faster connections now?
5. Whether such a solution can be reasonably scaled in the future. In terms of RAM or disks I know that there should be no problems, but what if you need to supply another server (or servers to keep an odd number of them). Would a mesh cluster still be a good solution in such a case?
Thanks for any feedback
Bartek
