[20:31:10] herron: hey, looking at that `service:error_slo:percent` mismatch for etcd -- do you think it's better to change the recording rule to be a 0-1 scale, or divide it by 100 in that prom query? [20:31:38] my instinct is it's better to always use unit scale under the hood and only ever render percentages for display at the last possible second, but interested in what you think [20:31:54] (and also no idea if it's too late given we already have that recording rule in place :P) [20:32:19] https://gerrit.wikimedia.org/g/operations/puppet/+/production/modules/profile/files/thanos/recording_rules.yaml#22 [20:32:35] yeah I'm afraid it's too late, at least if we want to keep the current metrics history without headaches [20:33:48] hmm -- can we deprecate that one and set up `service:error_slo:fraction` alongside it? at least that sets us up for the right pattern wrt other services [20:34:45] sure, yeah good idea, I wonder if we should do something like that by default. as in go ahead and create both variants? [20:35:07] well, I'm hoping we won't ever need `:percent` for anything except historical data on the one we've got [20:35:43] better to standardize on one for sure [20:35:44] unless you can think of use cases where `:fraction` wouldn't work? [20:35:50] okay cool [20:37:31] thanks for taking a look, I'll send you patches for that later today [20:38:02] for sure, sounds like a plan!