|I'm a freelance, junior sys-admin who is currently tasked with setting up the hardware for a tech startup. We're hosting a software service and need a system solution that makes sense within our budget and requirements.
The ideal goal would be n + 1 design, high security, long-term high availability without breaking the bank. Budget is ~6k for initial round of investment towards gear only. Additionally funds will be allocated for a year of quarter rack space in a datacenter local to my location.
I'm a bit over my head with my current knowledgebase, and intend to bridge the gap with a lot of pre-planning over the next 60 days. I figure 60 days before ordering any gear, 30 days to have it all come in, and get it initially configured at the datacenter, then 90 days to build out our hosting interface and properly test the system before going production. 6 months would be nice, but 8 months is being allocated for project.
Okay, so onto the gear and setup:
(2) Cisco SG-300-10 Switches, 1 wired & active, the second rack-mounted ready for wiring in case of failure.
Primary <> Secondary
With heartbeat in load-balancing + fail-over mode
Primary <> Secondary
Configured in PFsense for failover
A 4:1 ratio of active to failover App/Data Servers. I think failover can be configured with the Nginx Load-Balancer else PFsense.
Initial deployment is the two firewalls, two LBs, and 4+1 App/Data servers, with expectations to grow more App/Data as demand increases.
The service runs on LEMP stack. A master to master MySQL link between each A/D and the failover exists on separate partitions, separate instances of MySQL. It continuously synchs database with each active A/D, ready for activation in the event a LB declares an active dead. Higher resources in the form of more memory, higher thread count CPU, and n(A/D) hard disk space is allotted for failover.
Additionally, I plan to use an Anycast DDos prevention serviced by my colocation provider. I am wondering what drawbacks there are to my model.
The first 3-4 months is designing and deploying the system, the next 3-4 is linking the system to the software with the developers so that we can auto-provision services. I plan to utilize scripts heavily for this.
modified 19-Mar-16 13:45pm.