It’s Hard to Stay Quiet Sometimes

Unified Click

I know I'm not typical - not in the least. I can get carried away at times, my passion can get the best of me... but it's because I'm driven to give the best I can in all that I choose to do, and it really bothers me when I can't do that. I mean really bothers me when I can't do that.

This morning I came into work to see that things weren't looking all that good for the latency on my data stream. In fact, it's worse than I've ever seen:

Latency Graphs

Twelve solid hours of off the chart latencies. And when I dug into the Capacity numbers it was clear what was going on... we are getting lots of back-pressure on publishing to the shared Kafka cluster, and that backed up the bolts feeding those, and so on, and so on...

My concern is that we're only retaining data on the source kafka cluster for 72 hours, so if this continues for too long, we're going to be in real trouble because we're soon going to be hitting the retention limit on the kafka cluster, and then we'll start losing data.

Not good.

Yet the solution is simple - give me three more boxes, put them on my existing kafka cluster, and put all the traffic back to mine. That's where it used to be - and until the number of clients got too high - it was perfectly fine. Add a few boxes for the additional client load, and it'll be fine again.

And that's what makes staying quiet very hard... It's politics. They are playing politics with the quality of a service that has my name on it. And that really irks me. I have the ability to deliver a great product, but it's politics that is causing me to be associated with a bad one.