Intentionally closing tabs to upgrade to Monterey;

• Why Every Software Engineering Interview Should Include Ops Questions – Charity is easily one of the top people writing about tech.
• How Walgreens’ sloppy Covid-19 test registration system exposed patient data – “Security by obscurity is an awful model for health records,” Sean O’Brien, the founder of Yale’s Privacy Lab, told Recode.
• 6 ways of system resilience

I’ve gone on about this before but have clarified my thoughts further recently. I tried to put this on twitter but the thread didn’t post correctly. So here is Adam’s Keynote Heuristic.

A keynote should be one of the following;

• Controversial – ‘x is dead’, ‘you are all doing x wrong’, etc. which shows how you are correct in saying it (if inflammatorily phrased) and then a call to action to flip it around to make it not so much. Using this approach I might do ‘Your problems can’t be solved with automation’ which would illustrate how most of the problems I got asked to solve with automation are actually people problems (or process, which really are also people) and then hypothesize on how to solve them. Or maybe ‘all the problems in automation have been solved’ (which is my current fun one) where I ridiculously generalize things into patterns we were using a decade ago and how they have been updated for today which leads to a call to action to overhaul the selenium docs.
• Story time – This is literally, ‘I have cool stories to tell.’ Typically this person is not from the domain that the event is for but whose stories can wrap up nicely into actionable things that can be applied to the attendees. ‘So there was this other time where I forgot to unstub the payment gateways…
• Something from left field – I find this one actually the most interesting and this is where someone takes a different field entirely and applies it to the audience of the event. I’ve done this for testing audiences around when I ‘fixed’ the dryer and coached lacrosse. The anecdotes need to be fun. I often try to snag the loot bags from conferences that are sharing facilities to get insights I can steal for this sort of talk.
• State of the Union – As the name implies, these keynotes reflect a bit on the past to set context for the future. Ideally, its a 1/3 past, 2/3 future split.

Too often keynotes are really just track talks with a personality/name-brand. Which doesn’t diminish their content, its just that they are misnamed.

• Draw a picture. It will help.

• Ulysses S. Grant was only 40 at Shiloh.
• I really dislike people saying ‘our’ or ‘my’ team when they are not a player or owner. Parents and fans don’t get to say ‘my’ team.
• Hard things are hard until they are easy. Risky things are risky until they are de-risked. The two are annoyingly related.
• I have no clue what is happening in the X-Men books.

• Hockey [ok, sports] parents are crazy
• It’s not the day after that hurts, it’s the day after the day after. (Snowboarding crash.)
• Keep track of your time while doing it, as trying to remember what you did (or learned…) after the fact rarely works out well.
• Vagrant based development environments are brilliant. Yes, yes, Docker ones are the new hotness…
• There is not the equivalent of -s -N for mysql once you are in the cli tool so either you get column names and grids for all queries, or you don’t.
• Test your reporting
• Then test it again.

• I still hate installing doors
• There is always more implicit knowledge than documentation. (Which resulted in me changing my laptop issuance policy of to ‘Issue once configured’ vs the current ‘Here is your laptop and instructions’. The next evolution will be central management of policy, etc.
• Black Mirror is overrated.
• Another reason to disable form autofill – Ad targeters are pulling data from your browser’s password manager. (See also the central management of policy bullet.)

• A reminder; it’s always a people problem
• A reminder; trust, but verify
• Things I can never remember;
  • show index from <table>
  • show full processlist
• The default polling interval for Laravel workers is 3 seconds. If you have certain queues that are more time sensitive, you need to adjust accordingly. But you can only go as low as 1 second. This is another known issue with Laravel that is important when you build larger apps beyond toys. File under ‘things I shouldn’t have to do’
• Sometimes, its not [just] a query that is misbehaving that is causing a problem, but is a query running in a loop. And you are one of only 5 or 6 people who can trigger it due to environmental stuff.
• Environment matters. Our current nasty performance problem will only be triggered in production.
• I have completely lost all comfort doing frontside turns on a snowboard

• created_at, which is a fields that is created when using timestamps in Laravel Models/Migrations, is not indexed by default. So any time you use the built-in latest() or oldest() functions on an Eloquent query, you are doing a full table scan. There was an issue raised, but of course got shut down by the maintainer. Seriously, it annoys the hell out of my how small scale the Laravel team thinks. Scaling wut? At least when DHH was creating Rails he had a huge application to build/maintain. What’s Laravel got? Oh, a bunch of ecosystem stuff…
• Observability throughout a distributed system is a pain. I’ve been trying to diagnose a performance bottleneck and tracking a request through 3 difference systems is a challenge. I need to somehow inject an id into incoming messages and flow it through. And messages that are originated in the system on their way out.
• I found the source of the bottleneck at least. It looked like it was in an Eloquent save() call. Buuuut, its in an Event Observer on the ‘created’ event. I so wanted to blame the framework…

• I’ve always heard rumour that AWS billing is a nightmare, and I sorta got a peek at that yesterday. Part of what I need to figure out is how to attribute our AWS spend to various projects (for tax purposes as well as project profitability) and since I was in CloudFront doing something I poked around a bit. Yeeaaaah, what a mess. The billing page just as requests by region. The CloudFront usage report just has stats and resource ids. But not tags. I’ve not spent much time with this, so I’m sure there is a way to get the per item spend based on the tag, but I suspect there will be some scripting in the future to figure that out rather than one convenient report I didn’t have to build myself.
• Speaking of AWS tags, they appear to be case sensitive, so ‘Client’ is different than ‘client’ which is a pain. There are very few places where case matters. Stop it.
• Now the reason I am using takes is in the vain hope that the ‘resource allocation tags’ feature of billing works as one might expect. Except, you have to go into the page and enable them. Which I have only just now done. (For both ‘Client’ and ‘client’.)

• The MySQL Timestamp format goes to a single second of accuracy, so when you pull records and order them by a timestamp column they will be returned in their creation order if the timestamp is the same. When creation order really matters, you have to do something like

  ORDER BY created_at DSC, id DSC

  order by created_at DSC, id DSC

• Consistency matters. But so do boundaries. Just because something is called ‘data’ internally, doesn’t mean it has to be externally. Especially if ‘info’ is a better name externally. What’s worse though is when you have used ‘info’ in one api response, but ‘data’ every where else. Worse still, when you are the person who introduced the inconsistency.
• AWS OpsWorks for Puppet Enterprise would make my life somewhat easier, but would also add $200 USD a month to our bill. A couple more deals and I’ll totally pull the trigger on this.