Scaling Meetings

Let me entertain you 

Scaling Meeting

The first several months I could not figure out what does it mean at all. Every Thursday, East Coast morning time, engineers get together on Microsoft Teams, the leader of the meeting opens a dashboard with a bunch of graphs, and everybody is just silently looking at these graphs.

Some of thes graphs consist of similarly looking graphs of Dirichlet function (it's the function that is 0 on rational numbers and 1 on irrational). A bunch of other graphs have the shape of saw (y=x-[x]). These graphs show memory status and garbage collection actions. Judging by my previous experience, this must be fixed, because they show stupid memory leaks, most probably in one of the caches.

But nobody fixes anything. They just click their tongues, and that's it. At times they find peaks and study these peaks. Like physicists or astronomers: "wow, it's a supernova... or maybe just a solar activity..."

Suddenly it turns out that I'm also registered as a host of this show. Ok, and what am I supposed to do? You are supposed to know what to do! The link to this dashboard was sent to me by Nick, a nice guy; and I opened this dashboard, and scrolled it. Verbalized my view of these graphs: "here's a Dirichlet function, here's a fractional part function, and?" I asked them, but everyone's silent.

Eventually, suddenly, it turned out that I have to add a report into a special page on Atlassian. Well, that's not a problem for me. You want me to write something, I can write it easily. View me as a free-lance poet. I can dedicate a small poem to these curves, but...

And finally, finally! I was told what this shit is about and why we need it! All this means that if, suddenly, the processes start taking over more resources, we would need to do "scaling": request more processors from the upper management (and they'll cut our requests, the way they cut down our logs, in spite of all the SOX).

But we don't see any changes.

Last time I was such a show host it was yesterday. We had two non-trivial events.

Event 1. "And every evening at sun-down" a peak of solar activity Kafka activity (Kafka is what the whole world uses for passing around events). We discussed this event, and came to a conclusion that this is the time when cache is dumped into logs.
 

Event 2. On one of test machines one of the services had doubled the number of unused database connections. From two to four.

Here we had to go deeper to investigate. It happened October 8th, at 9:22 EST a connection was added. At 10:22 another connection was added.

Out of curiosity, we went to look for recent deployments. What did we deploy exactly. But well, Jenkins has so many deployments that we just couldn't find anything. Ok, and then I (as a show host) went to github to see the history of this service.

No code was changed for the last two years. But, October 8, in the morning, one build bot had upgraded a version of a library, which refers another library.

That's when the investigation was closed. I recorded it all in a doc on Atlassian, first having to figure out whether Atlassian Wiki even allows adding a row at the top of the table. Yes, there is this feature, if one googles it.

Now tell me, how much patience does it take to live with all this. And how long?

I'm starting to hate it, frankly.

 

читаю немецкую инструкцию к яйцеварке

 Инструкция на французском.

На первой странице изложено, как её разворачивать, как расправлять провод, и отправляют на седьмую страницу, где написано, как мыть яйцеварку, потому что её, как заключённого, прибывающего в новый лагерь, надо помыть.

Через две страницы - описание, как варить яйца, чтобы все сварились в одинаковом режиме (там есть ещё "неодинаковый режим").

На картинке, справа от ручки, управления нарисованы яйца - белое, чёрно-белое, чёрное. Соответствует выбору "a la coque", мягкое, твёрдое. На яйцеварке же, чёрного цвета, картинки в негативе - чёрное, бело-чёрное, белое. Важна позиция.

Это первая часть балета. Используется, если нужно варить семь яиц среднего размера, только что вынутых из холодильника, и имеющих температуру 7ªC.

Вторая часть - если вы варите меньше семи яиц или яйца мелкие. Слева от ручки есть два режима, тоже "чёрное" и "чёрно-белое" (на яйцеварке негатив). Эти два режима в рамочке, чтоб не перепутали. 

Третья часть - между режимами в рамочке и режимами без рамочки - там две точки. Первая используется, когда семь яиц среднего размера только что вынуты из холодильника. Вторая - если яйца крупного размера.

(Дальше идёт инструкция, как доставать их из яйцеварки.)

В реале же она запищала через минуту, яйца ни хрена не сварились, пришлось варить ещё раз.

И это только первые четыре страницы. Буду дальше читать.

На следующих написано, как варить, если несколько нужно в одной степени сваренности, а другие в другой. (Отгадка: поварили, выключили, часть вынули, остальные дальше варим.)

Там ещё четыре страницы, про то, как мыть (по-немецки), как снимать кальций (лимонным соком), как в ресайклинг (ресикляж) правильно сувать. Написано, что ресайклинг этого прибора - бесплатный. Ресайклинг тары - тоже бесплатный.

 

dybr

 Last night I wrote to Mike that I'm a developer, and not happy with the things I've been doing lately, and we should think about it and talk, e.g. Monday. Mike answered that this is my "user support" week, as agreed. And I wrote "Well, ok." And wrote Sean, my recruiter (I'd say, my agent).

Sure Sean will be happy to get me another job (and get paid for that); he wrote that he has something in mind, and we agreed to chat over the phone next week.

Then we had a "standup", where they started discussing this issue with an account being created five times (they call it "duplicate"). Kind of agreed to what we had agreed yesterday: that a user creates another account after getting an error - now the authorship of this idea is totally ascribed to Devon. After like me talking about it for a couple of weeks. See, I'm like an invisible middle-aged woman here now. I know, I know, I could have been more visible. But this is not my style. I don't fucking care. And no, I'm not angry, I just watch the show.

Eventually they came to a conclusion that an error happens because one of the services started adding one more field, "typeOfUser" to the XML they send. And you know, the whole fucking dumb universe validates the XML they get. And this field is "unknown". Is not it an idiocracy? Short, they just ignore records that contain an extra field. 

As a result, Devon, in a hurry, wrote an important update to our service: a debug log entry when an extra field is present. Good that I'm not included in his list of reviewers.

Oh, as to reviewing, Esteban last night fixed his "monadic" code (I posted it), so everything's fine now (and a test case is added), so I was glad to approve it. Well, we are professionals, are not we?

What next: I just shut the computer and... we went for a walk, and... eventually I opened the computer, looked around. There's a Dependabot notification that an Avro library has a security hole, and we need to update the dependency. I had to ask Sam (because she is the one that knows everything) about where is that dependency; she pointed it, and... and I updated that shit, and started running tests... except that one test did not even compile (yes, it's from master).

But anyway. Till Monday.