[08:09:30] <kormat>	 volans: is there an accepted approach to adding unit tests to python files in the puppet repo?
[08:12:02] <volans>	 kormat: I'm not yet around but the quick answer is take the code out of puppet in its own project :-P that said we have few examples, none of which are great IMHO, can link them in a bit
[08:16:03] <kormat>	 klausman: remember that issue with os.sendfile() and sys.stdout? it's stopped happening on my machine. w.t.f.
[08:18:21] <kormat>	 i haven't rebooted, i haven't updated python since before i ran into this issue
[08:21:00] <kormat>	 huh. some of my terminals have stdout opened with O_APPEND, and some don't.
[08:21:34] <kormat>	 yeah, that determines it. i've no idea why the difference.
[08:23:05] <kormat>	 https://phabricator.wikimedia.org/P12562#70292 ssh changes the flags...
[08:27:46] <klausman>	 weeeird
[08:28:52] <klausman>	 But it does something similar here
[08:29:00] <klausman>	 Before: flags: 0100002
[08:29:06] <klausman>	 After: flags: 0102002
[08:30:31] <kormat>	 it only seems to happen about 50% of the time for me
[08:31:29] <klausman>	 Hang on. This might not be SSH
[08:31:58] <klausman>	 I just tried producing the change 4-5 times in a row without success. Then I used *tab completion* once, and it appeared
[08:32:29] <kormat>	 yesss. that's it.
[08:32:37] <klausman>	 though not on, e.g. ls.
[08:32:57] <kormat>	 so maybe the host tab completion does something funky
[08:33:07] <klausman>	 Havin' a look-see
[08:33:25] <kormat>	 ping cum<tab> also does it
[08:33:40] <klausman>	 I suspect it's what ever looks in known_hosts
[08:37:34] <klausman>	 tmpkh=($(awk 'sub("^[ \t]*([Gg][Ll][Oo][Bb][Aa][Ll]|[Uu][Ss][Ee][Rr])[Kk][Nn][Oo][Ww][Nn][Hh][Oo][Ss][Tt][Ss][Ff][Ii][Ll][Ee][ \t]+", "") { print $0 }' "${config[@]}" | sort -u))
[08:37:36] <klausman>	 Lovely.
[08:40:22] <_joe_>	 my eyes!!!!
[08:41:03] <_joe_>	 kormat: look at what was done for mcrouter's certificate verification script, I would say
[08:46:51] <kormat>	 _joe_: ack, will do.
[08:58:38] <klausman>	 kormat: fwiw, by "breaking" my terminal using tabcomp on a hostname, I can make your python script fail as well
[08:58:43] <apergos>	 omg
[08:59:06] <kormat>	 klausman: good good :)
[08:59:36] * apergos checks right back out of this channel
[08:59:41] <klausman>	 But I have been entirely unable to find anything in the bashcomp function used that would explain thus
[09:02:12] <klausman>	 Now installing zsh to see if it has the same behavior
[09:03:24] <_joe_>	 ahah this is golden
[09:03:41] <_joe_>	 "python script broken by tab completion"
[09:03:49] <kormat>	 :D
[09:04:45] <klausman>	 zsh does *not* do this
[09:05:59] <klausman>	 This might be a *readline* issue, not bash
[09:07:07] <_joe_>	 would make sense
[09:07:22] <klausman>	 At any rate: WONTFIX UPSTREAM :-P
[09:07:49] <_joe_>	 definitely
[09:08:12] <_joe_>	 but I think kormat can verify the stdin flags and set them appropriately in her script
[09:08:49] <kormat>	 _joe_: i could, but what i've done is just open /dev/stdout and use that
[09:09:21] <kormat>	 rather than modifying the terminal flags, which seems naughty
[09:09:41] <klausman>	 Well, bash/readline futzing with them like it does is naughty as well
[09:09:51] <kormat>	 klausman: indeed
[09:10:08] <klausman>	 But I do get if you don't have the energy to pester upstream about it
[09:10:31] <kormat>	 figuring out which upstream is already non-trivial :)
[09:12:00] <_joe_>	 yes
[09:12:56] <klausman>	 https://sourceware.org/bugzilla/show_bug.cgi?id=14292 One of my fave bugs I ever reported. Bug in glibc starting in 2001, I found it in 2012, it got fixed in 2015
[09:13:17] <klausman>	 (and the fix was a literally a 1-character change)
[09:16:02] <kormat>	 klausman: it's awk
[09:16:19] <klausman>	 Oh. So the line up there even is the culprit!
[09:16:28] <kormat>	 yep
[09:17:21] <kormat>	 running `awk '' /etc/ssh/ssh_config` is enough
[09:18:31] <klausman>	 ioctl(1, TCGETS, {B38400 opost isig icanon echo ...}) = 0
[09:18:43] <klausman>	 Terminals. They will be the end of us all.
[09:22:32] <kormat>	 https://git.savannah.gnu.org/cgit/gawk.git/tree/io.c?h=gawk-5.0.1&id=ef83f3a14e5d3d2c446bdbdb801ba14345fc27ed#n983
[09:24:10] <klausman>	 Why, tho/
[09:25:14] <volans>	 kormat: I came back to elaborate on the above question, but I see you're doing a deep dive into much more important problems :-P
[09:25:52] <kormat>	 :D
[09:35:30] <volans>	 kormat: so we have a way to run per-module tox on puppet
[09:35:43] <volans>	 and so yes it's possible, but it's a bit of a mess
[09:36:00] <volans>	 and if a project grows more than one file it should really go into its own repo
[09:36:32] <kormat>	 volans: the example i'm thinking of is my binary-search script. it should only be a single file, but it would be nice to have some tests
[09:36:42] <kormat>	 and having a separate repo for it + packaging seems like way overkill
[09:36:48] <volans>	 indeed
[09:37:59] <volans>	 if you look at rake_modules/taskgen.rb (search for tox_ini)
[09:38:03] <volans>	 you can see what it does
[09:38:41] <volans>	 is your file in a module?
[09:38:50] <volans>	 or in profile
[09:39:08] <kormat>	 so far neither
[09:39:44] <kormat>	 i had originally started adding it to operations/software.git, but then realised that deploying from there is a nightmare
[09:50:08] <volans>	 yeah
[09:51:28] <kormat>	 i guess making a separate puppet module for this makes sense. it's a fairly general script, so it could be useful elsewhere in the future
[09:52:22] <_joe_>	 +1
[12:50:38] <marostegui>	 what's up with restbase?
[12:50:55] <kormat>	 it's restless(base)
[12:51:00] <volans>	 I was asking in -serviceops
[12:51:11] <volans>	 and the wikitech page is empty
[12:51:25] <marostegui>	 Did pdu work start?
[12:51:36] <volans>	 no idea
[12:51:45] <volans>	 didn't see any chat about them
[12:51:46] <volans>	 yet
[12:52:55] <volans>	 also why we're trying to get 'aggregated feed content for April 29, 2016' ?
[12:53:13] <moritzm>	 C4 and C5 don't list any restbase server
[12:53:13] <volans>	 a call to https://en.wikipedia.org/api/rest_v1/feed/featured/2020/09/15 seems to return some results
[12:53:21] <moritzm>	 (the ones scheduled for today)
[12:53:29] <elukey>	 so I am doing a roll restart of cassandra on aqs, and I see from the restbase grafana page that some 50x are returned
[12:53:32] <jbond42>	 i think this happened a month ago and there was an issue wit the alert fetching the wrong page
[12:53:34] <elukey>	 https://grafana.wikimedia.org/d/000000068/restbase?orgId=1&from=now-3h&to=now
[12:53:48] <elukey>	 I am not sure what feed featured is but maybe uses aqs
[12:54:02] <elukey>	 checking cassandra on aqs
[12:54:08] <elukey>	 (the cookbook is running)
[12:54:27] <marostegui>	 moritzm: today it is: C4, C5, C2 and C3
[12:54:33] <jbond42>	 recoveries comming in now
[12:54:48] <godog>	 I'm seeing some indexing error for logstash, that's aqs error logs for 'operation time out' talkign to cassandra FYI
[12:55:05] <godog>	 I'll followup in a task cc elukey
[12:55:21] <elukey>	 yes I think it is my roll restart, but not sure why
[12:55:25] <moritzm>	 marostegui: ah, yes, but also without restbase* in C2/C3
[12:55:35] <marostegui>	 cool
[12:57:38] <cdanis>	 kormat: I think it would be reasonable to make `doctest` work for Python in the puppet repo, fwiw
[12:58:20] <elukey>	 godog: I am trying to understand if wikifeeds uses aqs behind the scenes, if so the alarms are explained
[12:58:40] <elukey>	 but it is not great
[12:59:08] <elukey>	 (it would be not great that aqs not responding temporarily causes a storm of health check failures)
[13:00:11] <kormat>	 cdanis: so reasonable that you'll send a CR? :)
[13:00:20] <elukey>	 ah snap alarms again
[13:00:31] <cdanis>	 kormat: I don't know where you get such ideas
[13:01:07] <godog>	 elukey: not sure if related or not, the alerts I'm talking about are for malformed logs which can't be properly indexed, opened https://phabricator.wikimedia.org/T262920
[13:01:10] <kormat>	 cdanis: i should report myself to sobanski for being overly-reasonable
[13:01:32] <elukey>	 godog: ah ok sorry I thought you were referring to the storm of restbase alerts
[13:02:33] <sobanski>	 kormat: unacceptable
[13:02:42] <godog>	 elukey: np! but yeah maybe related, no idea atm
[13:02:44] <kormat>	 /o\
[13:04:22] <_joe_>	 elukey: so wait, wikifeeds calls aqs?
[13:04:34] <_joe_>	 why is that not in the config for wikifeeds?
[13:04:40] <_joe_>	 oh because it calls it via restbase
[13:04:45] <elukey>	 nono I have no idea, I am only saying that it matches
[13:04:46] * _joe_ cries in a corner
[13:04:52] <akosiaris>	 elukey: I don't think it does...
[13:05:05] <akosiaris>	 or... maybe I am wrong, let me make sure
[13:05:10] <elukey>	 see https://grafana.wikimedia.org/d/000000068/restbase?orgId=1&from=now-3h&to=now
[13:05:32] <akosiaris>	 it's the "most read" endpoint that is complaining
[13:05:38] <akosiaris>	 so this is starting to sound more plausible
[13:05:42] <elukey>	 yeah
[13:06:16] <elukey>	 I checked the spickerack logs and there are only two nodes left/in-progress to do, 2 instances each, the cookbook should finish soon-ish
[13:06:47] <elukey>	 maybe the defaults for the cookbook are a little aggressive for aqs
[13:06:51] <elukey>	 will investigate
[13:06:53] <_joe_>	 yes
[13:06:54] <akosiaris>	 this is /v1/page/most-read
[13:07:05] <akosiaris>	 so looks like AQS indeed
[13:07:14] <_joe_>	 I think restbase calls wikifeeds that calls restbase that calls aqs
[13:07:14] <akosiaris>	 making sure in restbase code/configuration
[13:07:16] <_joe_>	 ok
[13:07:25] <_joe_>	 akosiaris: wikifeeds has the same issue btw
[13:07:44] <akosiaris>	 damn I hate the interconnected mess that is restbase
[13:07:53] <_joe_>	 tell me about that
[13:08:35] <volans>	 I know I will regret to have asked it, but why the icinga alert reports 'aggregated feed content for April 29, 2016'?
[13:09:03] <akosiaris>	 it's probably the x-amples part of the spec
[13:09:08] <akosiaris>	 that date has been chosen
[13:09:09] <_joe_>	 volans: it's the description of the test in the spec
[13:09:12] <cdanis>	 _joe_: here, maybe this will help you feel better -- sound on 🔊 https://twitter.com/hitsugichan/status/1305700778695327744
[13:09:38] <volans>	 ack
[13:09:57] <volans>	 would be nice to add some info in https://wikitech.wikimedia.org/wiki/Services/Monitoring/restbase or a link any related existing docs fwiw
[13:10:47] <_joe_>	 I didn't even know that page existed
[13:11:12] <volans>	 I think are the auto-generated ones for generic alert
[13:11:15] <_joe_>	 and indeed
[13:11:18] <volans>	 IIRC Daniel worked on that
[13:11:24] <volans>	 don't recall all the details sorry
[13:11:26] <elukey>	 cdanis: I am playing it in a loop
[13:11:28] <_joe_>	 so yes, tell him.
[13:11:30] <akosiaris>	     if (req.params.domain === 'fy.wikipedia.org') {
[13:11:30] <akosiaris>	         return BBPromise.resolve({ meta: {} });
[13:11:30] <akosiaris>	     }
[13:11:36] <cdanis>	 elukey: <3
[13:11:38] <akosiaris>	 I am not sure I even want to know...
[13:12:47] <_joe_>	 elukey: so once the aqs reboot is over, please let us know
[13:12:54] <_joe_>	 so we can check if the issues continue
[13:13:05] <elukey>	 yeah it should finish soon, I am tailing logs
[13:13:13] <elukey>	 should be down to the last node
[13:13:24] <elukey>	 [DEBUG clustershell.py:254 in ev_pickup] node=aqs1009.eqiad.wmnet, command='c-foreach-restart -d 10 -a 20 -r 12'
[13:14:04] <akosiaris>	 I 've just verified that it's indeed AQS, wikifeeds wants to call the pageviews API
[13:14:10] <akosiaris>	 which is indeed AQS
[13:15:46] <_joe_>	 now, why can't wikifeeds call aqs directly?
[13:15:54] <_joe_>	 given aqs is basically restbase as well?
[13:17:34] <akosiaris>	 cause "it's the API"
[13:18:22] <elukey>	 cookbook done
[13:21:13] <_joe_>	 elukey: https://grafana.wikimedia.org/d/VTCkm29Wz/envoy-telemetry?viewPanel=6&orgId=1&var-datasource=codfw%20prometheus%2Fops&var-origin=restbase&var-origin_instance=All&var-destination=search-https_codfw&var-destination=wikifeeds
[13:21:28] <_joe_>	 I am so happy we have this kind of visibility now
[13:22:13] <elukey>	 nice
[13:25:21] <ottomata>	 hello!  am talkingg about stream processing stuff with some search team folks today
[13:25:37] <ottomata>	 i've mentioned potential SRE use cases before.
[13:25:40] <ottomata>	 q for all:
[13:25:42] <_joe_>	 ottomata: define "stream processing"
[13:25:45] <ottomata>	 ok
[13:25:50] <akosiaris>	 elukey: also https://grafana.wikimedia.org/d/35vIuGpZk/wikifeeds?viewPanel=20&orgId=1&var-dc=codfw%20prometheus%2Fk8s&var-service=wikifeeds
[13:26:11] <elukey>	 thanks!
[13:26:13] <_joe_>	 ottomata: if you mean having an infra that reacts to events, I think we should go with something native to k8s.
[13:26:16] <ottomata>	 https://en.wikipedia.org/wiki/Complex_event_processing
[13:26:34] <ottomata>	 and or
[13:26:34] <ottomata>	 https://en.wikipedia.org/wiki/Event_stream_processing
[13:26:38] <_joe_>	 an aws-lambda like model
[13:27:04] <ottomata>	 but more generally: if SRE could easily run SQL on streams in kafka, would that be usefful?
[13:27:12] <ottomata>	 for things like DDoS detection or the like?
[13:27:17] <_joe_>	 yes
[13:27:31] <_joe_>	 oh so you mean real-time processing of data in a stream
[13:27:35] <ottomata>	 _joe_:  def +1 to k8s stuff
[13:27:37] <_joe_>	 so flink
[13:27:58] <ottomata>	 flink or kafka streams or something, but probably flink given licensing issues with KSQL
[13:28:05] <_joe_>	 ottomata: for most stuff that's not analytics, I think something like knative or kubeless are probably the direction we want to go into
[13:28:40] <_joe_>	 if the pattern is "watch for events, react to the event" then something like that should be our choice
[13:28:50] <_joe_>	 or in the meantime our dear-old changeprop
[13:29:00] <ottomata>	 :)
[13:29:39] <ottomata>	 _joe_:  that sounds like just simple triggering on events though, right?
[13:29:48] <_joe_>	 more or less, yes
[13:29:58] <ottomata>	 not somethjing like: join these 3 streams on this key over 10 rolling seconds windows and look for this pattern and then alert
[13:30:03] <_joe_>	 not just triggering, ofc, but yes it's not stream /analysis/
[13:30:13] <_joe_>	 yeah def not
[13:30:32] <ottomata>	 is ^^ something that could be useful to SRE, if it was not painful to do?
[13:30:34] <_joe_>	 ok thanks I wanted to understand the boundaries
[13:30:43] <_joe_>	 it seems interesting to me
[13:30:45] <_joe_>	 cdanis: ^^
[13:30:57] <cdanis>	 yes, I am definitely interested
[13:31:19] <_joe_>	 ottomata: add us!
[13:31:20] <akosiaris>	 I am interested as well. As long as it's usable during an outage it could prove useful
[13:31:25] <ottomata>	 _joe_:  to meeting?
[13:34:49] <ottomata>	 would something like that also be useful for anything other than alerting?
[13:35:19] <cmjohnson1>	 FYI c3 one side going down now
[13:35:25] <cmjohnson1>	 I mean C2
[13:36:42] <cdanis>	 ottomata: potentially, yeah
[13:36:45] <_joe_>	 ottomata: I guess at some point yes
[13:37:50] <ottomata>	 ok, for concrete use case stuff, for the formal example of complex stream joining windowing alerting during an outage
[13:38:15] <ottomata>	 what's a more concrete way of stating that use case?
[13:38:30] <ottomata>	 like; during an outage suspected to be caused by a ddos
[13:39:53] <ottomata>	 it'd be nice to be able to write streaming sql that joins the webrequest stream with the netflow data by IP over a time window and emits events where count is greater than N ?
[13:40:06] <ottomata>	 (dunno if that is real, just made it up)
[13:40:47] <kormat>	 it's odd to see both "nice" and "sql" in the same sentence :)
[13:42:26] <_joe_>	 kormat: try sparql
[13:42:37] <_joe_>	 ottomata: we have several use cases
[13:42:48] <_joe_>	 one could be outlier detection and adaptive rate-limiting
[13:43:01] <kormat>	 _joe_: 😮
[13:43:11] <_joe_>	 kormat: or wait, try graphql
[13:46:17] <ottomata>	 hmm adaptive rate limiting interesting
[13:46:20] <ottomata>	 ok thanks all!
[14:18:44] <cmjohnson1>	 pulling power to C3 now
[14:21:28] <marostegui>	 cmjohnson1: c2 is fully done?
[14:22:08] <cmjohnson1>	 marostegui c2 power is fully restored, the pdu linking and setup is not done but that will not affect power
[14:22:57] <marostegui>	 cmjohnson1: excellent, thank you. Going to restart the services we have there. If you have a chance, db1088 is showing with one power supply as disconnected
[14:25:59] <godog>	 ms-be1024 and kafka-jumbo1004 too, same alerts
[14:28:50] <kormat>	 _joe_: any preferences on naming for this binary-search-of-ordered-logfiles script? i'm thinking `bsection`
[14:29:06] <_joe_>	  csection sounds better
[14:29:08] <_joe_>	 but ok
[14:29:43] <godog>	 🐝section
[14:29:44] <kormat>	 :)
[14:32:54] <_joe_>	 kormat: is your tool just selecting timeranges or it's a more generalized binary search tool?
[14:33:19] <kormat>	 _joe_: you give it an (ordered) prefix, it will output all lines that start with that prefix
[14:33:27] <kormat>	 it doesn't care about what the prefix contains
[14:33:29] <_joe_>	 ok
[14:33:40] <_joe_>	 so yes what godog said
[14:33:52] <_joe_>	 I dunno if utf-8 is allowed in puppet classes names
[14:33:59] <cdanis>	 only one way to find out
[14:34:20] <godog>	 hold my unicode, I'm going in
[14:35:49] <_joe_>	 thankfully it does not
[14:36:58] <kormat>	 _joe_: does the mcrouter tox env actually work? it seems to be telling me that `rm` is not installed
[14:37:25] <kormat>	 https://phabricator.wikimedia.org/P12590
[14:38:13] <volans>	 kormat: have you added rm to whitelist_externals ?
[14:38:20] <volans>	 that whitelist binaries outside of the venv
[14:38:34] <kormat>	 volans: i haven't touched the config at all
[14:38:40] <_joe_>	 kormat: it worked last time I used it for sure
[14:38:50] <_joe_>	 it caught a typo, but that was years ago
[14:38:50] <kormat>	 _joe_: what year was that?
[14:38:56] <_joe_>	 like 1 year ago at least
[14:39:12] <kormat>	 haha, i see
[14:40:19] <kormat>	 it looks like modules/envoyproxy contains a working python test env
[14:40:59] <_joe_>	 see? I didn't want to point you to my copy of what was done in mcrouter
[15:03:02] <cmjohnson1>	 marostegui: all power has been restored to c3
[15:03:53] <marostegui>	 cmjohnson1: thanks, db1090 reports one power supply disconnected, if you can double check? thank you!
[15:05:49] <cmjohnson1>	 moving on to C4 now, removing power from one side
[15:15:32] <godog>	 two hosts powered down in C4 now it looks like, ms-be1024 and thanos-be1003
[15:16:48] <godog>	 (back now)
[15:20:49] <cmjohnson1>	 godog both power supplies were plugged into side a for both servers
[15:21:41] <godog>	 heheh that'd explain, thanks cmjohnson1
[15:46:04] <cmjohnson1>	 marostegui: c4 all power restored...moving to c5
[15:46:26] <marostegui>	 cmjohnson1: thank you!
[16:35:19] <cmjohnson1>	 marostegui: power restored c5
[16:36:44] <marostegui>	 cmjohnson1: excellent,is there any other rack left?
[16:38:11] <cmjohnson1>	 done for today
[16:40:12] <marostegui>	 cmjohnson1: great, thank you! I will restart our mysqls then
[19:14:17] <mutante>	 I wish i could tell the puppet compiler to run on 'all hosts using a defined type' but that's not possible like it is with actual classes, or is it?
[20:44:50] <volans>	 you can get the list of hosts from cumin though :)
[20:46:48] <mutante>	 volans: good point, yes. thanks
[21:15:14] <sukhe>	 mutante: "I wish I could tell Puppet" t-shirt idea
[21:23:56] <mutante>	 sukhe: ;)
[21:24:41] <cdanis>	 yeah I've also written puppetdb queries for such, which is useful if you want to see all the params given to all instantiations of a certain class
[21:24:59] <cdanis>	 (recipe in a paste somewhere)
[21:32:26] <bd808>	 cdanis: is that equivalent to "the proof is too long to fit in this margin"? ;)
[21:32:54] <cdanis>	 I think it's secretly a gripe about Phabricator not having any sort of text (full-text or title-only) for pastes
[21:39:14] <chaomodus>	 nobody can tell puppet.
[21:42:25] <bd808>	 huh. I had never noticed before that the 'query' input in the "advanced search" for pastes is mostly useless. :/
[21:43:22] <bd808>	 It seems to a) only search the paste title and b) only return the first matching result
[21:45:40] <cdanis>	 yeah :(
[22:46:02] <mutante>	 cdanis: dunno why i keep telling them about the phab ticket..duh. maybe we need a backup ticket system outside our own infra
[22:46:10] <cdanis>	 ahah
[22:46:39] <cdanis>	 I just realized it'd definitely be possible to distribute tunnelencabulator as a Windows app
[22:47:44] <mutante>	 :)
[22:48:02] <mutante>	 maybe we should use translatewiki to get translations of our "how to report networking issues" page
[22:48:18] <mutante>	 let the pros handle it for all the langs
[22:53:30] <cdanis>	 I think we should work on the process some, before we do that
[22:53:41] <cdanis>	 I think a bunch there could be simplified
[22:54:12] <cdanis>	 it's pretty clear we have a bunch of users who look at the page, and have a reaction of "nope"
[22:56:35] <mutante>	 true
[22:57:43] <cdanis>	 I also have some ideas in my head about "there should be a simple executable that gathers this", but that's much harder ofc
[22:58:28] <mutante>	 the brandnew security.wikimedia.org website could have a drop form maybe
[22:58:37] <mutante>	 well, scratch that. i meant security-static :o
[22:59:01] <chaomodus>	 can somebody quick +1 this https://gerrit.wikimedia.org/r/c/operations/puppet/+/627627
[23:29:21] <chaomodus>	 thanks :)
[23:34:00] <cdanis>	 mutante: https://phabricator.wikimedia.org/T262613#6464857
[23:37:35] <mutante>	 cdanis: wow, seems very cool to use RIPE Atlas probes
[23:37:46] <cdanis>	 I don't know why I didn't try it earlier
[23:37:53] <cdanis>	 there were even more than I expected
[23:38:19] <cdanis>	 I'm not going to mess with it right now, but, tomorrow morning I'll add the stats I can (probably just the ping rtt) to our atlas_exporter, so we have some basic graphs
[23:38:30] <mutante>	 cool!
[23:38:30] <cdanis>	 unfortunately I don't think atlas_exporter groks traceroutes at all though