3 ms·

I was in the same boat for years, and just avoided offering any kind of hosting or SLA services. Eventually I just bit the bullet. And yes, it was exactly what

I was in the same boat for years, and just avoided offering any kind of hosting or SLA services. Eventually I just bit the bullet. And yes, it was exactly what you'd expect (waking up at 3am to fix problems, being available 24 hours, etc). That's the bad news. The good news is: These problems only exist for the first client who signs up for the SLA. Once you've got 2,3,4 clients retaining you, you can afford to farm the work out to others. But also, it just forces you to rethink how the systems are designed. Write more test coverage, build better failover mechanisms etc. Also, don't go for that "respond in 2 hours, fixed in 4 hours" crap. Most SLAs I've seen just state a guaranteed uptime % (99% or whatever) and you start prorating your retainer fee if the uptime slips below that. That way, if you really dont want to wake up (or if you're really unavailable to fix something) you don't have to, you'll just have to refund some of your retainer at the end of the month.

porker 12y ago
Very interesting, thank you. I agree about rethinking how systems are designed - I can see the need to build much more error-tolerant configurations. How did it work for you farming out the work? Getting up to speed on any codebase (such that you can fix issues under pressure) takes time, and to get people up-to-speed on 3-4 codebases sounds a big investment in them.
- MalcolmDiggs 12y ago
  Yeah, I never ended up letting 3rd parties loose in the code. I saved the actual problem-fixing for myself. I just hired some guys on Odesk to standby with a set of instructions on how to get the servers back up. Their only job is to keep the sites alive, and this is pretty easy because 99% of the crashes I experience can be band-aid'ed with simple restarts and such. So their instructions would say something like: Reboot the server If problem still exists: Do ABC If problem still exists Do XYZ ...stop when site comes back online. ...write up a crash report, send to me with the server logs When I get back on the clock I look over the logs and the report and take my time addressing whatever the underlying issue is.

porker 12y ago

Very interesting, thank you. I agree about rethinking how systems are designed - I can see the need to build much more error-tolerant configurations. How did it work for you farming out the work? Getting up to speed on any codebase (such that you can fix issues under pressure) takes time, and to get people up-to-speed on 3-4 codebases sounds a big investment in them.

MalcolmDiggs 12y ago
Yeah, I never ended up letting 3rd parties loose in the code. I saved the actual problem-fixing for myself. I just hired some guys on Odesk to standby with a set of instructions on how to get the servers back up. Their only job is to keep the sites alive, and this is pretty easy because 99% of the crashes I experience can be band-aid'ed with simple restarts and such. So their instructions would say something like: Reboot the server If problem still exists: Do ABC If problem still exists Do XYZ ...stop when site comes back online. ...write up a crash report, send to me with the server logs When I get back on the clock I look over the logs and the report and take my time addressing whatever the underlying issue is.