5 ms·

>As now we've introduced some peers code review, automatic testing on most critical stuff (but since the codebase sucks these aren't really reliable tests) The

>As now we've introduced some peers code review, automatic testing on most critical stuff (but since the codebase sucks these aren't really reliable tests) They may not be "reliable", but these are your safety net, or harness, so you don't fall. I wrote about similar issues, for instance here: https://news.ycombinator.com/item?id=26591067 https://news.ycombinator.com/item?id=26591067 and, given your promotion, here: https://news.ycombinator.com/item?id=37211796 https://news.ycombinator.com/item?id=37211796. It contains a few steps starting from "So...". You can add monitoring, something like Sentry (https://sentry.io https://sentry.io) will capture exceptions that were not handled that you have not seen because the stack trace is buried in hundreds of pages of logs or something. It groups them by exception and counts them. It's pretty awesome. (https://docs.sentry.io https://docs.sentry.io). It supports around 108 platforms (Java, Python, JavaScript, etc.). This lets you see the exceptions and makes prioritizing easier (which ones are the most frequent, which ones impact the most, etc.). If you don't have them already, issue templates are really useful and the comment I linked to explains why, but here's an example of an issue template (again, you can configure them for different types of issues so team members select from a dropdown for a bug or a feature):  ### Description ----------------  ### Target ----------  ### Instances -------------  ### Proposal ------------  ### Resources -------------  /label ~feature /cc @odle

odle 3y ago
thank you for your help, I'll implement what you suggested in the linked topics (improve the pipeline, code coverage and all the other management topics). About sentry, we've something similar, it's a manual integration with Elasticseach/Kibana, but it's shit. On the other hand, trying to use something like sentry instead, would look bad (it's about politics...)
- Jugurtha 3y ago
  I think you're using Elasticsearch/Kibana for logging. Sentry is mainly for exceptions and errors. You can self-host it: https://github.com/getsentry https://github.com/getsentry. The integration is probably fewer than 10 lines of code; it sits on top and you don't have to change how your code works or send things its way, Sentry does the capturing itself, so you wouldn't be using it instead of your E[?L]K stack, but in addition to it for exception monitoring. Not wanting to at least try it, or something similar to catch exceptions, especially when the current system isn't delivering, is what should look bad.

odle 3y ago

thank you for your help, I'll implement what you suggested in the linked topics (improve the pipeline, code coverage and all the other management topics). About sentry, we've something similar, it's a manual integration with Elasticseach/Kibana, but it's shit. On the other hand, trying to use something like sentry instead, would look bad (it's about politics...)

Jugurtha 3y ago
I think you're using Elasticsearch/Kibana for logging. Sentry is mainly for exceptions and errors. You can self-host it: https://github.com/getsentry https://github.com/getsentry. The integration is probably fewer than 10 lines of code; it sits on top and you don't have to change how your code works or send things its way, Sentry does the capturing itself, so you wouldn't be using it instead of your E[?L]K stack, but in addition to it for exception monitoring. Not wanting to at least try it, or something similar to catch exceptions, especially when the current system isn't delivering, is what should look bad.