|
|
|
|
|
by ibmthrowaway218
3768 days ago
|
|
> We also have to port our own servers to another service, and fast. Once you've done that you should look at what other services you rely significantly upon and see about mitigating the risk there by having an alternative ready to go (or even sharing the load). At a previous startup we realised we relied wholly upon Mandrill and so reimplemented the sending code so that half of the emails went out via Mandrill and the other half by SendGrid. A stunt like the above just requires a quick reconfigure to make all emails go via the alternative provider whilst we (with less panic) add another new alternative provider to share the load. It also helps build up a positive reputation before cutting over straight away. (This wasn't about splitting the emails amongst free tiers to keep it free, we were far away from moving up to a paid tier even with all emails going through one provider.) |
|
This also has the benefit that if one service goes down, we can automatically fallback to the other with (hopefully) no downtime. Most email providers have a limit to what you can send until you're established with them, so using another provider as a standard backup isn't feasible, as suddenly sending 1,000's of emails a day will see the account get suspended pretty quickly.
Plus, if one of them goes out of business or changes their terms with little notice, like Mandrill, we should have a bit more time to work around it as at least one service will work.
[1]: https://github.com/dchesterton/go-service