dybr

 I'm on pager duty today, so I was looking around, what's up, etc. There was one funny (or not) failure reported: a user sees "unknown error". But I could not get into datadog, because, well, it did not let me in. It did allow me to change the password, though. We have two datadog servers, on one I can get in, and on another I could not. Wrote to internal support, and they eventually posted a link, where to go - and kaboom, I'm in. Turned out, another man had the same problem, and he's in too.

That's how I saw this "unknown error" in the logs. But soon I saw a message that no, it's not our team's product that returns this shit. Ok, that was easy.

Bermi wrote to me some reasonable comments regarding the script. But well, the tests fail on master, so? Asked him what he suggests to do with failures on master.

What else? Oh, the bots pull requests. Some go through cool, but another was causing stack overflow in Scala compiler, mentioning scalameta, macros, java reflection, tpolecat's shit, etc. And the operation was "scalafix". So I experimented with it... it was not failing, even when I installed that "tumerin@11" or something JDK. So I wrote to the colleagues. Marc tried just to add more stack memory, and kaboom, it worked. Wow.

Then I turned to that "bad log" case, when an exception causes two dozen log messages, since every line goes separately to the log. Eventually figured that the cause may be in Vault class, in a library written by some other team. And the library we use is about a year older than the latest version. So I wrote to Morgen (the author of the latest version), asking if she thinks it's ok to update.

Also, that "kill persistence" case - Brad still has some ideas (good ones), so I'm tweaking it accordingly.

That's it. I would not say I spent tons of time working. Swimming, French, watering the lawns, figuring out insurance, bank things... and buying AAPL right before it crashed. Should not I just retire from trading? Maybe.