←back to thread

A computer upgrade shut down BART

(www.bart.gov)
232 points ksajadi | 2 comments | 05 Sep 25 14:52 UTC | HN request time: 0.232s | source
Show context
bigmattystyles ◴[05 Sep 25 16:40 UTC] No.45140579[source]▶
Broke the read-only Friday rule…
replies(2): >>45140943 #>>45141386 #
jkingsman ◴[05 Sep 25 17:11 UTC] No.45140943[source]▶
I know this is a tongue in cheek casual comment, but this article is a really good and important counterpoint: https://charity.wtf/2019/05/01/friday-deploy-freezes-are-exa...
replies(10): >>45140998 #>>45141312 #>>45141402 #>>45141439 #>>45141454 #>>45141628 #>>45141856 #>>45142250 #>>45142827 #>>45147202 #
yacthing ◴[05 Sep 25 17:47 UTC] No.45141402[source]▶
This reads like someone who works on a small and simple system.

"Deploy on every commit" lmao

"Shipping software and running tests should be fast. Super fast. Minutes, tops." hahah

replies(3): >>45141484 #>>45142070 #>>45143985 #
sampullman ◴[05 Sep 25 17:54 UTC] No.45141484[source]▶
Deploy to what? Staging on every merged PR (commit to stg), and prod deploy on every commit to main? That sounds reasonable to me, and I've done some variation of it on most projects for the last 10 years or so without issue.
replies(1): >>45141654 #
1. yacthing ◴[05 Sep 25 18:08 UTC] No.45141654[source]▶
Well people aren't talking about not deploying to staging on Fridays.

And there are hints to what the author actually means, like "Each deploy should be owned by the developer who made the code changes."

That just isn't feasible in a system that's of any reasonable size.

replies(1): >>45142051 #
2. da_chicken ◴[05 Sep 25 18:38 UTC] No.45142051[source]▶
Yeah, what happens when Team A makes a change and Team B makes a different, seemingly unrelated change, and they both get merged and pushed... only to have a dozen customers discover that if someone is using Feature X that Team A just worked on and Feature Y that Team B just worked on while they have Uncommon Option Q enabled, then their backend process server will crash taking down their entire instance.

Who's fault is that?

Asking because I have been the customer with Uncommon Option Q enabled.