A new tool for online development, AWS outage, and AWS' interpretation of DevOps.

HyperDev

Fog Creek, makers of among others Trello, showed off their latest project: an online editor/hosting service/everything else named HyperDev. The idea behind this project is that it gives you a complete and ready development stack to work with, without you needing my to do anything.

It starts really simple. You go to the HyperDev site, and that’s it. A small server is spun up for you and an example site is already running on it. Next up is just editing everything in real time, the moment you were it it will be executed. Which is very important to note while you’re writing potentially destructive code. So don’t just write DELETE * from users with the intention of adding the rest of the line. Your users table will be empty before you can do so.

Another interesting thing about it is the ability to invite others to work with you on a project. You can send someone an invite for your project and then you can collaborate in real time. I haven’t tried this out yet while playing with it, but from the examples in he site it looks pretty good. As it parses and shares updates in real time there shouldn’t be many conflicts either.

Obviously there are other limitations as well, it’s not really version controlled and the backend language is limited to NodeJS.

I feel that HyperDev has potential as a tool for creating quick mock ups and demos, but it seems a bit limited for more than that. It’s still in beta though, so there will be more improvements to come.

AWS Sydney outage

So, on Sunday the most important AWS region for my part of the world (aka Sydney) pretty much fell apart¹. Apparently this was caused by a big thunderstorm temporarily taking out a data center resulting in the loss of an availability zone.

Luckily this was eventually fixed, but for a long time people across Australia lost access to among others streaming services and food deal². Obviously a lot of these services weren’t limited to a single availability zone. A lot of the impacted services will have been set up so that they’d fall back to another AZ. Unfortunately this didn’t work well enough for some reason. Whether that reason was related to the sudden influx of new server requests or something else doesn’t really matter, but the results do.

The idea is that cross-AZ applications are more durable and should be able to withstand something like this. That it didn’t isn’t good, and I hope that they’re looking at ways to prevent this. As I write this I haven’t seen a post-mortem yet, so there might have been more going on. But, for anything that’s really important it’s probably highly recommended that you have at the least the ability to quickly spin up a copy of your system in a different region. The just released cross-region read-replicas of Aurora can probably help with this.

DevOps at AWS

In a more positive light, AWS released a dedicated DevOps section on their site. Obviously this is aimed at convincing you to use their services, but it gives a good overview and definition of DevOps in general and how what they offer fits in there.