16:00:56 <asn> #startmeeting SponsorR
16:00:56 <MeetBot> Meeting started Tue May 19 16:00:56 2015 UTC.  The chair is asn. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:00:56 <MeetBot> Useful Commands: #action #agreed #help #info #idea #link #topic.
16:01:04 <asn> hello friends.
16:01:09 <karsten> hi everyone
16:01:19 <asn> so who is here?
16:01:23 <nickm> me
16:01:29 <asn> no syversons or ohmygodels today.
16:01:36 <asn> but i think we have a full Tor house.
16:01:59 <asn> so let's start with status reports
16:02:04 <asn> i can start since i'm brief
16:02:28 <asn> during past week i was mainly doing exams and thesis. fortunately, the big load finished today. there is still more to come but that's in 7 days or so.
16:02:41 <asn> i talked with karsten adn dogulet about #15744
16:02:50 <asn> (which was just updated with graphs fromn what i see)
16:03:03 <karsten> 4 minutes ago!
16:03:04 <asn> and i also looked a bit at the new DoS #16052
16:03:29 <asn> also discussed with kernelcorn his SoP project.
16:03:34 <asn> anyway that's it from me.
16:03:36 <asn> who wants to follow?
16:03:57 <asn> karsten: go for it
16:03:57 <karsten> well, I could.
16:04:04 <asn> (seems like dgoulet is missing, but it's ok)
16:04:08 <karsten> https://people.torproject.org/~karsten/volatile/torperf-hs-on-ferrinii.png
16:04:15 * asn looks
16:04:24 <karsten> we now have torperfs set up to measure hidden service performance.
16:04:35 <karsten> they're running on ferrinii, a tor vm.
16:04:43 <asn> interesting
16:04:46 <asn> and this is on the real network
16:04:50 <karsten> they're writing tor's logs to disk, which is not ideal.
16:04:55 <karsten> yes, real network.
16:04:57 <asn> and they fetch every how many minutes?
16:05:19 <karsten> 51200 every 5 minutes, 1024*1024 every 30 minutes, 5*1024*1024 every hour.
16:05:33 <asn> right on
16:05:42 <asn> what do we hope to find from this?
16:05:43 <karsten> so, I need to write some code to get these measurements onto metrics.
16:05:52 <asn> or just get a baseline?
16:05:57 <karsten> baseline, I think.
16:06:04 <asn> of average bandwidth on hidden services?
16:06:07 <karsten> wasn't this one of our milestones for april or may or june?
16:06:09 <asn> *of average bandwidth of clients on hidden services?
16:06:20 <asn> karsten: prossibly.
16:06:28 <asn> great.
16:06:37 <karsten> so, this graph is just to show you that these measurements are happening.
16:06:40 <asn> ack
16:06:44 <karsten> events contain lots of details that are not shown here,
16:06:47 <asn> right
16:06:53 <karsten> including substeps of the connection establishment process.
16:07:18 <karsten> so, my goal is to write that additional code to get these measurements archived and onto metrics in the next weeks.
16:07:29 <asn> great
16:07:31 <karsten> so, that's one thing.
16:07:38 <karsten> https://trac.torproject.org/projects/tor/ticket/15744#comment:12 is the other thing.
16:07:43 * asn looks
16:07:44 <karsten> especially https://trac.torproject.org/projects/tor/attachment/ticket/15744/introduction-circuits-2015-05-19.pdf
16:08:03 <karsten> I can include more data from dgoulet and others there.
16:08:36 <asn> useful.
16:08:47 <asn> yes, getting more data would help.
16:08:56 <karsten> ok.
16:08:58 <asn> i will look at the graph and think some more of other visualiztions that we might want.
16:09:03 <asn> thanks :)
16:09:05 <karsten> great!
16:09:08 <karsten> that's all from me.
16:09:13 <asn> ack
16:09:22 <karsten> ah, maybe one thing:
16:09:35 <karsten> I learned that I'll have to spend quite some time fixing/migrating yatei soon.
16:09:38 <karsten> because it's dying.
16:09:47 <kernelcorn> Yawning: yes, so can chain .tor to .tor, which is quite powerful. I"m aware of loops, I have not yet addressed that
16:09:54 <karsten> I shouldn't commit to doing more R things this week, I think.
16:10:00 <asn> karsten: ok
16:10:03 <asn> karsten: no worries
16:10:16 <asn> if there is something *urgent* i wll tell you -- but i don't expect it to be
16:10:21 <karsten> cool!
16:10:26 <asn> so brief meeting today
16:10:29 <asn> do we want to talk about something?
16:10:32 <Yawning> kernelcorn: ok
16:10:36 <asn> i haven't had time to think much of sponsorr the past week
16:10:55 <asn> but it seems like we are on track  (?)
16:11:17 <karsten> I have no idea, to be honest. are we?
16:11:24 <karsten> we are doing things.
16:11:25 <kernelcorn> Yawning: also, if I control a.tor, I automatically have control over b.a.tor and c.b.a.tor, and so on.
16:11:42 <asn> karsten: i feel we are OK in terms of statistics for July.
16:11:49 <asn> karsten: with #15744 and #15513
16:11:50 <karsten> ok, great!
16:12:03 <asn> (and i have more data that we can graph on #15513, but i will wait to gather more till june)
16:12:24 <karsten> ah, that one. cool.
16:12:27 <asn> karsten: and i think dgoulet is going strong with his hsdir health.
16:12:31 <asn> so i would say we are ok.
16:12:39 <asn> i can speak some more with dgoulet about hsdir health later.
16:12:52 <karsten> sounds good!
16:12:55 <asn> karsten: did you see anything interesting on the graphs you made?
16:13:21 <karsten> well, 80% of introduction circuits not being used at all was surprising.
16:13:54 <karsten> and 5% having a lifetime over 25 hours.
16:14:01 <asn> is that 80% or 90%?
16:14:04 <asn> looking at the first graph
16:14:08 <asn> it's hard to distinguish between 0and 1
16:14:17 <dgoulet> I'm here!
16:14:25 <karsten> hi dgoulet
16:14:32 <dgoulet> so sorry was eating lunch and forgot about the new time :(
16:14:37 <asn> dgoulet: hello. prepare your status report and we will be with you in a sec.
16:14:38 <asn> dgoulet: no worries.
16:14:41 <karsten> asn: you mean 0 vs. 50?
16:14:52 <asn> karsten: yes correct.
16:14:57 <asn> 0 or 1-50
16:15:04 <asn> ok
16:15:07 <asn> that answered my question i guess
16:15:12 <karsten> 0 is from 0% to 80%, 1-50 to 90%.
16:16:23 <asn> ack
16:16:28 <asn> now it makes sense
16:16:45 <asn> i think also a simple histogram might be useful in this case.
16:17:11 <asn> but anyway, i will look into the graphs more later.
16:17:14 <asn> dgoulet: whatsup?
16:17:15 <karsten> for a presentation?
16:17:34 <dgoulet> ok
16:17:41 <dgoulet> Working on little-t tor SponsorR patches, most of them in need_review. HS Health improvement, I pushed a design doc (not amazing but a start) mostly made from my notes I wrote down on paper.
16:17:46 <dgoulet> Turns out the hs-health is detecting misbehaving HSDir as a side effect :). I've reported two of them on bad-relays@ last week. It's getting more and more stable so next step is definitely try to come up with useful visualization using the logs.
16:18:01 <karsten> (asn: talking about graphs on the ticket sounds fine.)
16:18:06 <asn> karsten: ack
16:18:10 <dgoulet> I might need karsten graph experience for that
16:18:11 <asn> dgoulet: where can we find the design doc?
16:18:17 <dgoulet> on the git repo
16:18:20 <asn> ok
16:18:29 <dgoulet> https://gitlab.com/hs-health/hs-health/blob/master/design.md
16:18:35 <karsten> dgoulet: sure, just let me know what you need.
16:18:38 <asn> ack
16:18:43 <asn> oh wow, not in gitweb.tpo.
16:18:44 <dgoulet> will improve it, some small features and future works are missing
16:18:50 <asn> dgoulet: ok
16:18:57 <asn> btw, sorry for not remembering,
16:19:06 <asn> but did the hshealth thing produce graphs in the past?
16:19:11 <asn> will it produce graphs for july?
16:19:16 * asn is all about those graphs
16:19:20 <dgoulet> this is turning into something bigger and bigger so before it gets to ugly code wise and feature wise, I need to take a step back and come up with a better design doc.
16:19:27 <asn> dgoulet: ack
16:19:50 <dgoulet> no the graph that were produce in the past are the tracing analysis, different
16:20:09 <asn> i wouldn't go overboards with super clean architecture and stuff. but yeah, making it less ugly is definitely good.
16:20:26 <asn> hoverboards
16:20:44 <karsten> clean design is good.
16:20:45 <dgoulet> imo, we can definitely graph "churn effect over time" that is when it happens and how much time it takes to recover
16:20:52 <karsten> you might have to maintain it for years..
16:20:57 <dgoulet> karsten: exactly
16:21:10 <asn> dgoulet: churn effect of what?
16:21:11 <dgoulet> it's pretty useful and I would like to integrate in the long run DonnchaC bad hsdir detection
16:21:20 <asn> dgoulet: hs descriptors?
16:21:25 * isabela around
16:21:28 <dgoulet> asn: when a HSDir is getting replaced by an other one for a desc ID
16:21:31 <isabela> sorry reading backlog
16:21:50 <dgoulet> asn: there is a delay before the descriptor shows up on the new HSDir because of different consensus on the HS and client
16:22:33 <asn> aha
16:23:07 <asn> ok will be cehcking the design doc soon hopefully
16:23:42 <dgoulet> I want to send karsten the logs it has and see how we can visualize some of it
16:23:57 <karsten> dgoulet: please do.
16:23:57 <asn> feel free to cc me
16:24:08 * DonnchaC checks it out
16:24:09 <dgoulet> FYI, to run this tool you need upstream master, stuff that the tool needs has been added very recently
16:24:55 * dgoulet done report
16:24:57 <Yawning> (I guess I did R stuff >.>)
16:25:00 <asn> thx
16:25:06 <asn> dgoulet: do you have anythig you want to discuss?
16:25:07 <Yawning> (#16052 is R right?)
16:25:25 <asn> Yawning: maybe. it's more SponsorU i think
16:25:32 <Yawning> oh ok
16:25:35 <dgoulet> yes I do have something to discuss
16:25:41 <Yawning> *wanders off*
16:25:49 <asn> Yawning: https://trac.torproject.org/projects/tor/wiki/org/sponsors/SponsorU/Tor
16:25:54 <asn> Yawning: sponsoru seems to be the DoS one.
16:25:59 <asn> dgoulet: bring it on
16:26:01 <isabela> (let me know if you want my help for this -> we should figure out a plan to brainstorm the Right Guard Behavior this month.
16:27:02 <dgoulet> we need a 0.2.6 release because of #15801, HS health tool is seeing more and more timeout on HSDir on the first and second try, I had report from HS op (quite used one) that people are complaining more of rechability issue lately
16:27:11 <dgoulet> so I don't know if it's ^ specifically
16:27:26 <asn> nickm: ^
16:27:35 <dgoulet> but right now it's pretty common that we have a .onion that has one or two valid HSDir only
16:27:39 <Yawning> should I work faster on this mitigation hack?
16:27:47 <Yawning> so we can include a backport as well?
16:27:56 <dgoulet> and my guard is starting to show signs of stress with more and more warnings (the one running hs-health)
16:28:22 <asn> dgoulet: what does the guard have to do with this?
16:28:48 <nickm> anybody want to help with the next 026 release then?
16:28:55 <asn> nickm: what needs to be done?
16:29:08 <nickm> see the release checklist in doc/HACKING?
16:29:21 <nickm> is there anything there that could be done by a non-nickm? :)
16:29:28 <Yawning> (I shall take that as a no)
16:29:34 <dgoulet> asn: I don,t know, it's just something I've seen lately, maybe too many hsdir requests or ... we have .onion right now that have only one HSDir with like 80kb/s bw... so means everything goes there ...
16:30:03 <dgoulet> thus I think we might want to merge the "is_stable" patch for HSDir flag also in 0.2.6?
16:30:27 <asn> nickm: changelog could be curated by non-nickm
16:30:37 <asn> nickm: and testing i guess
16:30:49 <dgoulet> and maybe #15963 also
16:30:53 <nickm> testing would rock; I can do the changelog pretty fast for this one
16:31:17 <Yawning> fixed "bugs"
16:31:23 <asn> ok i will try to give it a bit of testing today or tomorrow.
16:31:25 <Yawning> *ducks* *runs* *hides*
16:32:06 <dgoulet> I also have another topic I would like to discuss
16:33:12 * dgoulet waiting for the current topic to be set and close :)
16:34:07 <asn> dgoulet: maybe previous topic is closed now
16:34:28 <dgoulet> yeah ok
16:34:31 <asn> with "nickm will try to release next 026 soon. need help in testing"
16:34:40 <dgoulet> good
16:34:46 <asn> dgoulet: are you saying that there are patches not merged in 026, that you want in the next realese?
16:35:02 <Yawning> Stable for HSDir
16:35:07 <Yawning> I think isn't merged
16:35:10 <asn> #8243?
16:35:14 <asn> ok
16:35:18 <dgoulet> yeah this ^
16:35:23 <dgoulet> and maybe #15963?
16:35:33 <asn> ok will try to look soon
16:35:40 <nickm> I would like as little backported as possible
16:35:53 <nickm> and only things with a lot of testing
16:35:59 <nickm> please keep that in mind?
16:36:00 <Yawning> I stand by my rationale for #8243, but
16:36:43 <asn> dgoulet: #8243 and #15963 are not as urgent as #15801, right?
16:36:50 <asn> they are just "would be nice if they were in"
16:36:51 <dgoulet> 15801 is already merged
16:36:51 <asn> or not?
16:36:55 <asn> yes
16:36:56 <dgoulet> well 15850 sorry
16:37:31 <dgoulet> yeah but Yawning has a rationale for #8243 to be merged in 0.2.6 so "should be consider" is what I would say
16:37:39 <Yawning> I think changes to the voting there should happen in the 0.2.6.x timeframe, but I don't view it as urgent
16:37:57 <Yawning> like, not "oh my god there's only 1 HSDir for foo.onion" sort of urgent anyway
16:38:08 <asn> ack that's what i think too
16:38:15 <asn> dgoulet: next topic?
16:38:16 <dgoulet> makes sense
16:38:18 <dgoulet> ok next topic
16:38:20 * asn needs a break
16:38:27 <dgoulet> (will be fast)
16:39:17 <nickm> can somebody put all that sutff in an email to me?
16:39:37 <dgoulet> actually nvm, nothing important now :P sorry
16:39:42 <asn> nickm: i can do so
16:39:56 <asn> except if dgoulet wants to, since he knows these tickets better currently.
16:40:34 <asn> dgoulet: you sure? ask away if you want!
16:40:39 <dgoulet> well it's pretty easy, if we think #8243 can wait, well 0.2.6 branch is ready for a release
16:41:00 <asn> right
16:41:01 <dgoulet> and DirAuth must deploy it :)
16:41:08 <dgoulet> else no point of a release :)
16:41:13 <Yawning> imo let the changes bake a bit, also do the fast thing, and do it next release?
16:41:18 <Yawning> when things aren't quite as on fire?
16:41:42 <asn> Yawning: you mean do #8243 and fast thing on next release?
16:41:44 <asn> i'm fine with this.
16:41:47 <Yawning> yah
16:41:49 <dgoulet> +1
16:41:50 <asn> ok
16:42:13 <asn> in that case, we just need a 026 release. and work on #8243 and fast thing soon, so that it's ready for next release.
16:42:19 <asn> that simplkifies things.
16:42:20 <Yawning> yes
16:42:35 <asn> right on
16:42:39 <Yawning> (my rationale for deploying them in a stable branch is correct right?)
16:42:56 <dgoulet> ok cool
16:43:29 <asn> Yawning: you mean so that dirauths can deploy them eaiser?
16:43:32 <asn> makes sense
16:43:35 <Yawning> yah
16:43:38 <asn> ok i guess this sums it up
16:43:42 <asn> i will send a brief email to nick
16:43:49 <asn> with the summary of what we jsut said
16:44:04 <dgoulet> awesome
16:44:17 <asn> ok thanks for the meeting guys
16:44:24 <nickm> thanks all
16:44:25 <asn> i will take a small break now, and be back in an hour or so.
16:44:30 <dgoulet> thanks!
16:44:31 <asn> i will be around irc this week.
16:44:33 <karsten> thanks!
16:44:39 <asn> so i will be easier to reach :)
16:44:40 <asn> take it easy!
16:44:41 <dgoulet> you guys got my email about this week right?
16:44:42 <asn> #endmeeting