I saw some chatter in slack this morning about folks not being able to resolve DNS for www.yahoo.com, so I started doing a little poking around. My first visit was to Downdetector, which looked a little like this:
Okay, so people on the Internet noticed it. …but did we (meaning LinkedIn) notice it? There may be other places where we leverage Yahoo APIs, but the one that jumped most immediately to mind was contact imports. Once Upon a Time I probably would’ve had the dashboard memorized, but I haven’t worked in the Growth space for a little while so it took me a bit of grepping around:
metrics -fg prod list -s zeppelin > zeppelin.metrics && grep -i yahoo !$
That turned up a list of pretty likely candidates - things with names like:
zeppelin/zeppelin.i001.Zeppelin_Provider_Sensor.Zeppelin_Provider_Stats.yahoo_ERROR.rrd
zeppelin/zeppelin.i001.Zeppelin_Provider_Sensor.Zeppelin_Provider_Stats.yahoo_SUCCESS.rrd
So what did those look like?
Yep. As expected, errors went up and successes went down (in fact, went to zero) in close alignment with that timeframe.
So I do still remember how to inGraphs! Coolcoolcool.