16:16:27 <hiro> #startmeeting network-health 2024-05-06
16:16:27 <MeetBot> Meeting started Mon May  6 16:16:27 2024 UTC.  The chair is hiro. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:16:27 <MeetBot> Useful Commands: #action #agreed #help #info #idea #link #topic.
16:17:00 <hiro> here is the usual meeting pad: https://pad.riseup.net/p/tor-nethealthteam-2023-keep
16:17:09 <GeKo> o/
16:18:06 <hiro> ok nothing bold on the pad good
16:18:23 <hiro> I have a little psa i am out next week, so I was wondering what you wanted to do with the meeting?
16:18:55 <GeKo> no strong opinion on my side
16:19:04 <GeKo> we could just cancel it or someone else runs it
16:19:08 <GeKo> *shrugs*
16:19:15 <hiro> yeah up to you and @juga
16:19:22 <juga> same as GeKo
16:19:28 <hiro> what do you prefer? for me it is ok to cancel and we sync in lisbon
16:19:42 <GeKo> cancel
16:19:46 <juga> ok
16:19:56 <hiro> sounds good then, no meeting next week
16:19:57 <GeKo> leaves time for other work :)
16:20:23 <hiro> great!
16:21:13 <ggus> o/
16:21:20 <hiro> hello @ggus
16:21:30 <hiro> meetup is on sat right?
16:21:37 <hiro> 1900 utc?
16:21:58 <hiro> is there anything needed from our side? the agenda looks good
16:22:20 <ggus> yes
16:22:50 <ggus> the agenda; https://pad.riseup.net/p/tor-relay-may24-meetup-keep
16:22:58 <ggus> i'll ask pavel to promote on social media
16:23:23 <hiro> nice
16:24:40 <ggus> do we have any updates regarding conflux performance?
16:24:56 <hiro> regarding graphs you mean
16:24:57 <hiro> ?
16:25:03 <ggus> yes
16:25:11 <hiro> I am still pondering if it makes sense to do extra graphs
16:25:17 <hiro> I have mixed feelings
16:25:52 <hiro> because I do not want to give the idea that we have to defend that
16:25:58 <hiro> what do you think @GeKo ?
16:26:23 <hiro> we have graphs on the website... I could just make them bigger
16:26:27 <ggus> and it seems that some operators have disabled conflux on their exits (or i had that impression by reading some emails on tor-relays). should we verify that?
16:26:35 <hiro> so these are easier to read
16:26:49 <hiro> I have asked directly ggus and it seems this was somethign temporary
16:27:06 <GeKo> i am not a fan of extra graphs
16:27:32 <GeKo> pointing to onionperf stats on metrics.tpo should be enough
16:27:52 <GeKo> iirc they show how conflux improved throughput
16:28:51 <hiro> yep
16:29:22 <GeKo> https://metrics.torproject.org/onionperf-throughput.html
16:29:49 <GeKo> https://metrics.torproject.org/onionperf-throughput.png?start=2024-02-06&end=2024-05-06&server=public
16:30:00 <GeKo> just compare the conflux with the non-conflux variants
16:30:48 <GeKo> we or look at the csv powering that graph and import that into whatever graphing you wanna do
16:31:04 <GeKo> s/we//
16:32:17 <hiro> yeah I think I agree...
16:32:33 <hiro> my doubt comes from the fact that the graph is all squashed
16:32:56 <hiro> and I wanted to download the csv and put it in the presentation gus does at the meetup so that people can read it properly
16:32:57 <ggus> did we already share this graph on that mail thread about disabling conflux?
16:33:22 <hiro> no it wasn't shared in the thread but it was mentioned at the last meetup
16:34:20 <GeKo> hiro: i think i tis kind of squashed as the de-conflux seems to be busted
16:34:27 <GeKo> which is probably a thing we should fix?
16:34:35 <hiro> yeah for public downloads
16:34:39 <hiro> I have no idea why
16:34:43 <GeKo> it just gives 0s back
16:34:53 * GeKo the onionperf saga continues...
16:34:56 <hiro> it uses the same configuration and firewall rules as op-de
16:35:49 <hiro> I have even recreated the machien and run on the side... but something is not working with iptables anymore...
16:36:25 <hiro> and I wonder why
16:36:29 <hiro> it is only with aws
16:36:38 <hiro> and on that datacenter for de
16:36:44 <GeKo> if you look at the .png it's not too hard to check out the details for the hk8 and us8 cases, just ignore the de8 one
16:38:05 <hiro> yeah I think that works too, let's see how we can frame this @ggus
16:39:24 <GeKo> oh, i guess that's https://gitlab.torproject.org/tpo/network-health/metrics/onionperf/-/issues/40072 as well
16:39:35 <hiro> yep I have it for this week
16:40:24 <GeKo> very weird
16:40:41 <hiro> yes...
16:41:04 <hiro> anyways
16:41:59 <hiro> is there anything else @ggus?
16:42:35 <ggus> yes, do we have any tool to check if an exit has disabled conflux?
16:42:59 <GeKo> hrm
16:43:37 <GeKo> not that i know of
16:44:41 <juga> is that published in any of the relay's descriptor documents?
16:44:43 <ggus> https://lists.torproject.org/pipermail/tor-relays/2024-April/021595.html
16:44:49 <ggus> > This tor is a relay and ConfluxEnabled is set to 0. We would ask you
16:44:49 <ggus> > to please write to us on tor-relay at lists.torproject.org or file a bug
16:44:49 <ggus> > explaining why you have disabled this option. Without news from you,
16:44:50 <GeKo> i am not sure whether we can easily detect that anyway. dgoulet might know
16:44:50 <ggus> > we might end up marking your relay as a BadExit.
16:45:02 <ggus> we should add this to the bad relays criteria wiki page
16:45:06 <GeKo> yeah, i was thinkin about that
16:45:11 <GeKo> i did that already
16:45:18 <hiro> +1
16:45:42 <GeKo> juga: well, the capability is published in the proto line
16:45:56 <GeKo> but i think that one is not affected by torrc changes
16:46:06 <juga> GeKo: ic, thanks
16:48:37 <ggus> GeKo: nice!
16:49:46 <GeKo> hiro: if you need something from me for the exception reporting framework lemme know
16:50:16 <GeKo> are we supposed to ack the estimations gaba shared? i think i forgot about that part...
16:50:30 <hiro> I thought that was an example ?
16:50:42 <hiro> I didn't check that yet
16:51:25 <hiro> I wanted to ask you if you had time tomorrow or wed to double check that with me... I think some of the activities we can use similar estimation to the other grant application we did that wasn't accepted
16:51:38 <GeKo> let's do wed
16:51:50 <GeKo> i think i need the time tomorrow to prep the s112 sync
16:52:02 <hiro> sounds good
16:52:26 <hiro> which reminds me, we have s112 sync one hour earlier
16:52:34 <hiro> 1500 utc
16:53:22 <GeKo> yep!
16:54:22 <hiro> ok nice!
16:54:34 <hiro> is everyone good?
16:54:40 * juga is
16:54:43 * GeKo is
16:54:49 * hiro is too
16:54:51 <ggus> hiro: i might not be able to attend the meeting tomorrow. but i can update the pad
16:54:58 <hiro> ok thanks @ggus
16:55:18 <GeKo> ggus: sounds good.
16:55:29 <GeKo> might be worth sync at some point, though
16:55:52 <GeKo> maybe lisbon would be enough assuming i 'll make it
16:56:06 * hiro hopes so
16:56:18 * juga too
16:56:19 <ggus> in 2weeks /o\
16:56:42 <GeKo> yeah, time is flying :)
16:56:48 <hiro> yeah
16:57:39 <hiro> ggus: quick question do we have a documentation sync on thu?
16:58:28 <ggus> hiro: i'm not sure
16:58:31 <ggus> gaba: do you know?
16:59:10 * GeKo gotta go ttyl!
16:59:18 <hiro> byeeee
16:59:19 <juga> nc says so
16:59:28 <hiro> anyways we can check later
16:59:35 <hiro> let's close the meeting
16:59:38 <hiro> #endmeeting