Align on use of AP labels
Following discussion came up in https://gitlab.com/gitlab-org/gitlab-ce/issues/37292
"how urgent is this? + how many people will this impact?"
I think the urgency still needs to have a big consideration in the labeling, otherwise the labels lose relevance because there's a lot that one can dream up that will impact a lot of people, but may not be urgent or immediately required.
@mydigitalself per https://about.gitlab.com/handbook/engineering/performance/#performance-labels, the urgency has the following example metrics:
Urgency: Examples
- U1
- Outage likely within a month.
- Affects many team members and/or many GitLab.com users
- U2
- Outage likely within three months.
- Affects some team members and/or a few GitLab.com users
- U3
- Outage can happen, but not likely in next three months.
- Affects some team members but no GitLab.com users
An urgency score and an impact score together set the AP label. Would you like to discuss?
As I've said before, for me an ~AP1 should be of similar impact to an ~SL1
where it's practically drop everything you are doing now and sort this out because we can't continue to function without this getting fixed, the ~AP2's - for the most part - get scheduled within 1 or 2 releases and if there's bandwidth, we deal with the ~AP3's. That's pretty much the guidance I'm looking for. As you've said before, of course, prioritisation is up to me but it seems wrong to be providing guidance that doesn't reflect how I should prioritise.
In the case of this issue, if it was an ~AP2 it still means it's very important and we will look to address it soon, given the resource constraints we currently have in ~Platform.
@mydigitalself I'm uncomfortable with the idea that the person picking the label has to worry about the effect the label will have, since that takes the objectivity out of setting the label. Trying to provide you with as objective as possible information to facilitate but not take over your scheduling role.
I'm fine changing the urgency and impact metrics to a degree where the objective picking ends up aligning more closely with what you'd expect the effect to be. For example, we can move the "number of people affected" from urgency to impact, and we can add "Performance substantially degraded (> 500 ms in many cases)" as an urgency U1, similar to the "outage likely within a month".