search

konflux-ci / architecture

Technical and Architecture documents
https://konflux-ci.dev/architecture/
18 stars 70 forks source link

chore(RHTAPWATCH-359): create ADR for availability probe framework #118

Closed yftacherzog closed 1 year ago

yftacherzog commented 1 year ago

This change defines the convention according to which appstudio components' availability will be reported.

yftacherzog commented 1 year ago

@avi-biton @amisstea please take a look.

eisraeli commented 1 year ago

Very informative. LGTM.

yftacherzog commented 1 year ago

@eisraeli @amisstea I added the points mentioned on our call.

yftacherzog commented 1 year ago

I feel like there are a lot of references with words like availability and metric. Are all uses of these words consistent? Can we define contexts for these words in the relevant places?

If there is reuse of a common word for two applications, we can try to draw attention to the different applications of the words using techniques like bolding the word groups.

@arewm I'll look into that. metric for sure often have multiple meanings in the context of monitoring. I'll try to see if there are points here which might be confusing in that sense. Regarding availability, we state that it's up for the individual teams to define what available means for what they report availability for.

yftacherzog commented 1 year ago

@arewm I tried amending the text according to all your comments. Let me know your thoughts.

yftacherzog commented 1 year ago

@ralphbean @adambkaplan Can I have your reviews on this PR? Andrew was reviewing it and we had some discussions going, but I'm hoping to have the PR approved by someone outside the o11y team even before Andrew is back.

yftacherzog commented 1 year ago

@adambkaplan

My first question regarding this ADR - why can't ArgoCD do this for us? Is this a missing feature, or something that we currently aren't doing that makes it harder for ArgoCD to figure out that our applications are synced AND operational?

the main aspect here is granularity I suppose. The idea is to provide a mechanism to allow us to probe components from the outside and verify that a certain behavior is as expected. I'm not sure this is something we can do with ArgoCD.

More to the point - would we ask a public user to implement the same kinds of availability probes for their workloads running on RHTAP?

Those probes are just for the availability of the services that constitute RHTAP. i.e. our own services.

yftacherzog commented 1 year ago

Hi @arewm, can we merge this?