Hire part time or temporary first, then full time.

Get a phone with push notifications. Setup a second system to constantly check for failure in the primary system. Get the second system to notify you in case of failures. Better yet, if you can, have the second system be a failover clone and have the notification dashboard be a third system. Base all these systems on different vendors systems if you can.

Buy 3g/laptop/other mobile ssh solution and use this to fix things when you're more than 30 minutes from your office/home.

And lastly, don't worry so much. Money isn't everything, and you honestly sound like you're possibly going to worry so much about this, it will make your life a net negative.

If this is the case, don't worry so much. If your app uses a well-understood (by you), proven (by others) software stack and you've been testing well enough, your worries may be for naught.

Some folks around here recommend tools like pingdom. Get something like that if it helps you enjoy a break with your friends.

Good luck!

Not in a single-person startup, but there were some useful HN-linked articles in the past. In particular, I found patio11's article about running a software startup as a single-founder (and on 5 hours a week) particularly useful: http://news.ycombinator.com/item?id=1206649

I think he's made some really good points about doing a lot of time-as-asset stuff (go read the article, it's really good). Maybe you should get in contact with him?

I wish you the best of luck.

Not sure exactly how he set it up (possibly through Twilio as well - http://www.kalzumeus.com/2009/12/29/twilio-phone-call-web-ap...) but his phone calls him playing Ride of the Valkyries anytime the site goes down.

1) Look into chef to automate 99% of your sys-admin tasks (setup, tear up, tear down, deployment)

2) Set up pingdom for monitoring your servers externally

3) Set up hoptoad (http://www.hoptoadapp.com) to monitor for exceptions and errors internally with e-mail notifications

4) Set up a laptop with a MiFi for remote access - you're going to keep this in your car whereever you go (if you're in a city or something, set up VNC on an iPad or something and have access to a workstation where you can look into emergencies)

The above will a) notify you of critical things happening and b) give you a way to act upon them. Thats pretty much all that you can do at this point -- and I'd look into Rackspace managed services to see if you can leverage them for basic system troubleshooting (this is on my ToDo list).

By the way, we've got a single founder community going, its new and we're still getting to know eachother but it is great for moral support - let me know if you want to join.

Just do whatever it takes for you to fall asleep at night. I recently bought a smartphone (N900) that allows me to ssh into my server from anywhere, so if a small fix is needed, I can do that quickly. Before I had that, I actually took my laptop camping with me during a busy period last year and drove to a nearby coffee shop periodically to check for any important issues.

Other than that, I'd say to scale up your architecture as your site gets more popular. If you later need some dual-DB setup with a heartbeat monitoring system then you can add that when the time comes.

Have good backups too. That might be more important than just uptime.

1) How do you monitor your site to alert you of problems?

2) How do you prevent your site from going down in the first place?

In my opinion, #1 is more straightforward than #2, and the more uptime you need for #2, the more you're going to pay.

a) Write a script to send you an email / SMS when your site is down. Run it on a separate server, or run it on something like Google App Engine.

I'm sure there are some simple free tools out there to achieve this for you. A quick search provided:

http://www.siteuptime.com/

They offer a check per 30 minutes for free. Can't beat free, but this is not an endorsement, I've never used them.

b) Use something with more features. Run it on a separate server or use someone who provides it SaaS style.

Nagios, ZenOSS, Zabbix, Groundwork, OpenNMS. I can personally vouch for Nagios. It's pretty easy to get a simple configuration going, and then it can get very complicated (you might want to monitor the monitor, right?).

2) If you've got two servers that you can connect to a load balancer, you may be able to run in active-active mode so that if one fails, you simply lose 50% capacity. For your data store, options include a database solution like MySQL in master-master form, or a relational database that has data redundancy as a feature. In my opinion you don't choose your data backend solely because of it's redundancy and failover capabilities, but it could be a factor.

If you can't do active-active mode right now for some reason, then active-passive can allow you to stay up enough to deliver that message until you restore your services. If you can get an "extra" IP address, you can even do this without a load balancer involved. Take a look at keepalived.org to see how you can float an IP address between multiple servers. CDNs such as Akamai also provide site failover features that might be worth investigating.

There are so many ways to skin the cat on this one, and I'm just scratching the surface. If your site is more than informational (e.g. if you're building a web based service or application), then monitoring and failover / redundancy are critical to your success. If you just want your informational site to be up, it's still pretty important. The fact that you're thinking about this now and not after your first 24 hour window of downtime is a good sign!

If you provide more information about what your startup is doing (at least from an tech architecture perspective) and what sorts of resources you're willing to spend to improve your uptime and failover capabilities, you might get more specific suggestions.

Best of luck, and congrats on the startup!

Pete

It ran smoothly for months, and then we had a bit of a problem, but site canary alerted us right away.

Also, site canary was recently made free, so set it up! It will email you when you have an issue. Should definitely have some sort of uptime tool like pingdom or site canary.

Disclaimer: I have nothing to do with site canary, other than I use their software.

As for automating restart, are you running on AWS? Then you can automate that. For that part of your question, you have to give some technical background on your setup etc...

The first thing to accept is that this is by definition an irrisponsiable thing to do...;)

Then you need to flow with it. The main thing is not aiming at fixing everything as fast as possible but rather in a way that will be as reliable as possible. Or in other words, your primary consideration should not be an early alert on Friday night, but rather a wilingness to work the weekend making sure whatever happened will not happen again.

The tech details really depend on the specific application.