Re: Using AWS ephemeral SSD storage for production databaseworkload?

On Mon, 29 Jan 2018 23:27:32 +0530
Pritam Barhate <pritambarhate@gmail.com> wrote:

> In short, I am just trying to learn from other people's experience.

This is identical to solutions that use tmpfs on linux for 
database storage or dealing with a fully failed storage 
system. Think about what you'd do if a RAID controller 
fried and botchd your entire array at once. You'll feel
just the same way if a box using ephemeral storage goes
down.

Your application needs to handle restarting transactions
and either a reverse proxy/load-balancer or client-side
switchover.

Depending on your tolerance for data loss you might need
three servers up, on as a secondary failover if the primary
fails so that you (pretty much) always have two servers up
to maintain the data. The last server only has to last long
enough for a restart and recovery so it might not have to
be very heavy duty, it's main purpose is to keep the database
alive long enough to recover the "real" server.

Q: Why not just use RDS?

It'll be simpler.

-- 
Steven Lembark                                       1505 National Ave
Workhorse Compuing                                 Rockford, IL 61103
lembark@wrkhors.com                                    +1 888 359 3508