[08:12:41] swfrench-wmf, sukhe: the answer is https://wikitech.wikimedia.org/wiki/DNS/Discovery#Failure_scenario ;) [08:15:31] <_joe_> volans|off: nope [08:15:31] if done via spicerack there is a check to prevent that btw: [08:15:31] https://gerrit.wikimedia.org/r/plugins/gitiles/operations/software/spicerack/+/refs/heads/master/spicerack/dnsdisc.py#262 [08:15:57] Idon't know if the confd template has a similar one [08:36:10] oncallers: I'm going to do some config tweaks to benthos@webrequest_live as part of T369737 [08:36:11] T369737: Site Issue: Delayed data in the `webrequest_sampled_live` kafka topic - https://phabricator.wikimedia.org/T369737 [12:12:39] hi folks anybody knows what's the best way of detecting that a golang program exhausted file descriptors attempting to establish a TCP connection? [12:13:15] rrors.Is(err, os.ErrPermission) works as expected to detect if for some reason the program hasn't got privileges to do so [12:13:19] *errors.Is [12:30:31] ok... errors.Is(err, syscall.EMFILE) captures it as expected [12:43:37] <_joe_> vgutierrez: sorry I missed it, but yeah I would've said os.ErrPermission so I would've been useless :) [12:44:13] yep, I need ErrPermission to detect setsockopt() failing to set a proper SOMARK [12:44:29] if for some reason the PID doesn't at the very least CAP_NET_ADMIN privileges [12:44:36] *doesn't have [12:45:16] better that than getting "down/ko" results in healthchecks without further explanation :_) [14:35:01] v.olans: thanks for highlighting that! yeah, those docs initially had me thinking failoid was used as a fallback for active-active too. however, I was recently looking at our gdnsd configs for unrelated reasons, and could not see how that might be the case :) [14:35:55] also yeah, there's no confd check to prevent depooling both on a discovery service AFAIK - just the check for preventing multiple pooled on an active-passive service [15:49:43] pws update-keyring is failing for me in the pw repo, is that a known issue? [15:50:12] (to be more specific, pws is failing with 'ed' too, which is what I started with) [15:51:16] seems like usually when I have this problem mutante sorts me out :)