Week 22: tcp nodelay linux networking, aws bedrock ai audio summary, okta architecture documentation, jenkins casc ansible, production richardcook incidentresponse, metrics dashboards monitoring

• It’s always TCP_NODELAY. Every damn time: Interesting low level kernel networking behaviour that can cause trouble to distributed systems builders. (I kind of am I guess and so is anyone who has to operate more than a single process.) I do believe I work at a layer above where this becomes an issue though? At least I don’t believe I’ve ever seen strange latencies in the work I do.
• Hacking our way to better team meetings: Tiny app doing summarization of an audio file on aws with bedrock.
• How we do Architecture at Okta - Monica Bajaj, VP Engineering & Mark Voelker, VP Architecture: Talks about leveling up tech and people. Architects are hands-on. Big emphasis on sharing knowledge and learning and they do that by writing everything they do down.

• geerlingguy ansible-role-jenkins: An ansible role that configures a jenkins server. Plugin management isn’t something that the casc plugin can do by default for whatever reason. Jeff has figured out a way using a jenkins provived cli tool (api client) and api. Neat.
• jenkins plugin Configuration as Code: Since I mentioned casc above, here’s a link. Casc is able to manage tools, the jenkins server environment, credentials, and likely a few more things I haven’t gotten into yet.
• The problem with a root cause is that it explains too much: Subtle point here but our attempt to look for a root cause can close our minds to understanding the bigger picture. There are many vulnerabilities (risky aspects that given the right conditions could lead to an outage but that generally don’t) in systems and we can’t dig into them all. Business priorities often differ from what we believe needs attention right now.
• The error term isn’t Pareto distributed: Mindset shift here too away from looking for major incident contributors (eg “20% of services cause 80% of incidents”) to looking at broader context with an open mind and learning.
• There are multiple uses for metrics (and collecting metrics): Why we collect metrics and often “overcollect” things we don’t necessarily have a need for right this minute for an alert.