[00:45:36] I'm trying to figure out a somewhat performant way of querying edit sizes (rev_len - rev_parent_id's rev_len) via Quarry. My intuition was to try to use a left join from revision against itself on rp.rev_id=r.rev_parent_id, but that seems to never return / timeout, reven if run against a single explicit rev_id and a LIMIT 1. [00:46:08] https://quarry.wmcloud.org/query/18750 [00:52:44] Krinkle: I think there might be something weird with that particular revision's parent revision (827022456). SELECT * from revision where rev_id = 827022456; is really slow [00:53:27] it was a random edit I picked from yc ontribs, https://en.wikipedia.org/w/index.php?title=Fruitcake&diff=next&oldid=827022456 [00:54:30] sorry, nevermind, i had a typo in my query [00:54:55] I'm trying a different approach now with a subquery https://quarry.wmcloud.org/query/22240 [00:55:26] overall, what I'm trying to do is return for a specific user name, all edits that are either page creations with rev_len > X or diff > X, where X is some threshold. [00:55:35] That's super weird though, i don't see any reason for that query to be slow [00:58:27] It also seems like its only slow on quarry and not toolforge replicas [00:58:42] web or analytics cluster? [00:58:55] quarry's on the analytics cluster, but otherwise they share the same replicas [00:59:54] it might be that specific host tho... [01:00:32] 17REPLICA_DOMAIN: 'analytics.db.svc.wikimedia.cloud' [01:01:13] > BIGINT UNSIGNED value is out of range in '`enwiki`.`revision`.`rev_len` - `enwiki`.`revision`.`rev_len`' [01:01:15] darn [01:02:12] https://phabricator.wikimedia.org/T178848 well, I'm not the first, that's always nice [01:02:29] I could have sworn I tried to search phab for a task mentioning a query like that, but maybe I searched wrongly [01:05:21] ok, so sub queries are weird in select, I was expecting that to be weird given nobody recommends it, and indeed it's not really what it seems (for one, you can't do where clauses on the result of those). So back to joins it is, but need to figure out why they're not working. [01:08:07] I tried using plan `revision` instea of `revision_userindex` but that didn't help either. [01:16:15] ok, well, this subquery hackery is satisfactory for now. attempt 3: https://quarry.wmcloud.org/query/18753 [01:16:46] using HAVING allows the subquery to be filtered [01:17:03] I considered doing the where clause inside the subquery, but then I can't tell the difference between new pages and small edits [23:56:55] What's the "Ack" button button do on a review comment at e.g., https://gerrit.wikimedia.org/r/c/operations/puppet/+/523667/2..4/modules/profile/manifests/prometheus/alerts.pp#193 [23:57:14] I assume "acknowledge" but I feel it could also be "ack, I didn't notice that!" [23:58:31] acknowledge for sure [23:59:49] weirdly it's not mentioned at all in the docs, so it must be