It’s actually quite surprising that we were able to get away without a rewrite for so long. As Flowdock became more popular and especially when support for multiple flows per window was added, this became one of our biggest architectural pain points. We tracked who was online by having clients send regular pings to the server, which were then relayed to everyone else in the flow. When Flowdock was launched, it was a web app that had one flow per browser window. Here’s a recap of what we learned while tracking down the bugs. Very quickly, our public Streaming API started running out of memory and needed to be constantly restarted. It’s what powers, among other things, our push notifications and Streaming API, and has been pretty solid for both purposes.Ībout a month ago, we rolled out major changes to our online status tracking service. Node.js is a big part of the Flowdock infrastructure.
0 Comments
Leave a Reply. |