SadServers

SadServers

@sadservers.com

SadServers.com - Practical Linux & DevOps Troubleshooting

433 followers89 following354 posts

"This is way more interesting than learning from tutorials. Finally found a good learning platform for me to learn about Linux servers and DevOps skills to further improve my skill set and career opportunity!" love.sadservers.com/t/7acad25d-8...

love.sadservers.comTestimonial from VinceThis is way more interesting than learning from tutorials. Finally found a good learning platform for me to learn about Linux servers and DevOps skills to further improve my skill set and career oppor...

(Funny they included the cents in the figure). So Linux sysadmin / DevOps / Infra Eng / Cloud Eng / Platform Eng / SRE / flavour of the month whatever want to call it Eng will still be in demand πŸ™‚

We'd rather spend time adding tangible value to our users (new scenarios, new features) than doing maintenance. BUT it can be argued a security issue -even if unlikely - could affect users, so we upgrade to keep our bar of no dependencies with Critical or High vulnerabilities.

It's been 35 years since Linus announced his Linux hobby project. It's a light one but I'm surprised not a lot of people have heard about his book talking about the early years. Same with Kernighan's book, which is a great read and anybody in the Unix/Linux world should read imho.

Have we invented a better command dependency tool (DAG) than "make"? it's from the 70s and I use it all the time (my cli is all 'git' and 'make' with the occasional 'docker')

"overnight success" πŸ˜€ SadServers has been updated every week in the last two years, either code or scenarios (weekends especially productive, one gap due to external emergency)

Called this back in May. Copilot has different price tiers now with a high-cost one coming up.

SadServers@sadservers.com

(maybe naive) GitHub outages seem to come from load and they have a built-in agent (Copilot) that easily writes a lot of code and Actions tests. Possible mitigation is making Copilot work less by 1) making it more expensive 2) capping its use and 3) tweaking it to be less verbose

Using Sentry for incident, logs and metrics. It's way cheaper than DataDog and the cost can be anticipated (you can cap it too).

If the AI support agent of a multibillion company like @stripe hallucinates a dashboard option and highlights in a statement the opposite of reality, imagine what the AI agents doing support on other sites are saying. (Unlike other sites at least with Stripe I can RTFM)