Ask HN: Turning my SaaS into an Automata?
As a solo founder I often wonder how long my SaaS application could survive without me. How long would it take my customers to notice I had been struck by the proverbial bus?
I haven't had to reboot a single of my server/app nodes in months although I do occasionally suffer downtime to one of my nodes due to DigitalOcean issues.
My database backups are automated, my logs rotate before I run out of disk space and hosting payments are on autopay.
I do not however have autoscaling or any service which automatically reboots instances if they go down...if they do.
What are all the little things I'd need to do to turn my SaaS application into a full Automata?
Something that could survive decades without any human intervention.
7 comments
[ 3.0 ms ] story [ 23.7 ms ] threadGoing on vacation is always a nerve wrecking endeavor since I don't want to by checking my health stats every day and have to scramble to a shell while on the beach (pun almost intended).
In some ways, building a fully self maintaining application is almost more important while I'm still here.
I'd be very interested to hear what everyone's most common maintenance tasks are they are performed manually.
For example, has anyone written anything to auto detect database corruption and automatically restore DBs from the latest backup? It's rare, I know, but if you want to run for decades without human intervention, it needs to be automated.
The company was sold over a year ago and seems to have had near zero work for a year before that. It seems they dropped the price and left it idle.
There's a calendar invite app on the Google apps for enterprise marketplace that's similar. They however officially placed it in maintenance mode rather than "just stopped talking"
I guess all these touches such as displaying the latest tweets or automatically displaying system status all makes it look very active and alive.
The copyrights still say 2012.
The idea is that if anything ever goes wrong, you fix it immediately. But then you figure out what you can change in the software so that that particular thing (or class of things) never happens again.
This isn't just for software errors. It works for customer support as well. If you keep getting mail asking the same question, it's a sign that something's not clear enough in your onboarding or documentation. If people keep asking for an account tweak, you need to either make that into something they can do themselves or, if they already can, do a better job of surfacing the feature so that the next guy will find it.
Over time, your inbox will shift away from panicked customers in need of an immediate response, and towards "thank you" type mails and feature requests, with the occasional oddball request that actually takes a bit of your time.
Sprinkle all the above with a bit of "setting customer expectations" so that your users know ahead of time that two days is a reasonable amount of time to wait for a response for an email.
That won't necessarily get you decades off, but it will let you crew a sailboat across the Atlantic with a reasonable expectation that your business will still exist when you pop back onto the grid a month later.
That's great advice. I have done this to a small extent with my onboarding and making sure my conversion funnel works, but I need to be more religious about it.
I can definitely see how this approach would lead to a very solid product.