IRC office hours/Office hours 2014-01-23
The office hours on 2014-01-23 were about migrating community tools from the Toolserver to the Tool Labs. Topics covered:
- migrating OpenStreetMap (18:02–19:02);
- migrating datacenters — bugzilla:55929 (18:41–18:46);
- implementing crosswiki joins (18:50–18:52);
- prioritising tickets for migration (18:52–18:55);
- Java servlets — bugzilla:54845 (18:55–19:00);
- migrating @toolserver.org email addresses & domain (19:00–19:33).
All times are in UTC.
[17:58:13] <Silke_WMDE> Coren: you there?
[17:58:37] <Coren> Silke_WMDE: I am.
[17:58:44] <Silke_WMDE> hey there!
[17:58:48] Coren waves.
[17:58:56] <Silke_WMDE> What are the nicks of Alex and Tomasz?
[17:59:05] <Krenair> ... Alex who?
[17:59:07] <akosiaris> me
[17:59:13] <tfinc> Silke_WMDE: i'm right here :) <--tomasz
[17:59:15] Krenair is also alex
[17:59:22] <Silke_WMDE> ah hi akosiaris tfinc! :)
[17:59:39] <Silke_WMDE> Waiting for Sumana...
[18:00:15] <tfinc> Silke_WMDE: last i saw her she was heading over to our offsite location for the architecture summit. i bet she's spinning things up there now
[18:00:27] <tfinc> i chose to say in the office for this to have fewer distractions
[18:00:37] <Coren> Silke_WMDE: akosiaris and tomaszf
[18:00:38] <Krenair> Sumana is here at SPUR
[18:01:03] <Silke_WMDE> So let's start, okay?
[18:01:22] <Krenair> over the other side of the room. and we're about to start, so I can't really go and talk to her...
[18:01:49] <Coren> We can start, it's logged anyways so she can catch up.
[18:02:23] <Kolossos> Hello
[18:02:32] <Silke_WMDE> I'm curious about OSM - what are your plans?
[18:02:38] <Silke_WMDE> Hi Kolossos
[18:02:52] <tfinc> I can jump in for that
[18:03:05] <tfinc> but i'll need akosiaris help for the ops bits
[18:03:13] <akosiaris> :-)
[18:03:29] <tfinc> I've done an inventory of OSM tools running on TS from the various wiki pages about
[18:03:39] <tfinc> i need help vetting if i've missed anything. whos up for reviewing it ?
[18:04:03] <scfc_de> tfinc: Put it up on a wiki?!
[18:04:30] <tfinc> i can easily put it up on MW.org just need one or two people to say the'll jump in so that i can move with that quick. you in scfc_de ?
[18:04:46] <Silke_WMDE> MAybe Kolossos?
[18:05:20] <Kolossos> I took a look to the inventory and it looks ok.
[18:05:39] <Kolossos> https://docs.google.com/spreadsheet/ccc?key=0AlEIL4N8KjNndHlsTXoxeFhGWmR4YS1JLU1PN3BGZXc#gid=0
[18:05:57] <tfinc> yup thats the one
[18:06:10] <scfc_de> I think if Kolossos okayed it, it's fine.
[18:06:16] <tfinc> that gives a nice list or at least a first pass of whats active, where it lives, and what we think needs to move
[18:06:18] <akosiaris> what are the numbers ?
[18:06:20] <tfinc> an action list for us
[18:06:51] <tfinc> akosiaris: their references from the wiki page they came from. a really long list of everything that was and/or might have been on TS
[18:07:02] <tfinc> i had to make my own list to make any sense of it
[18:07:09] <akosiaris> oohhh ok
[18:07:11] <tfinc> so now that we have this in place we know what we have to move
[18:07:11] <akosiaris> thanks
[18:08:05] <tfinc> on the ops side i've spoken mostly with brandon and i defer all of the expertise to akosiaris but what i was told was that the current specced hardware can easily take the existing TS load but is very short on hard drive space
[18:08:12] <tfinc> as in we barely have enough for raid redundancy
[18:08:19] <tfinc> but i'm sure akosiaris knows more
[18:09:11] <akosiaris> not that much on that front. There have been some ideas about putting the tiles on Ceph but well ceph....
[18:09:46] <Silke_WMDE> to all: If you want me to collect your questions, send them in a query
[18:09:47] <tfinc> akosiaris: so is it a file system, physical drive, etc issue ?
[18:10:33] <akosiaris> More like a where to cache/store the tiles issue I think
[18:10:55] <tfinc> right thats what i thought
[18:10:59] <tfinc> so thats a blocker for us right now
[18:11:17] <Silke_WMDE> Kolossos: Do you know how that's done on TS?
[18:11:21] <akosiaris> well there is an idea to cache them into varnish but I have not yet come to implement it
[18:11:38] <nosy1> currently tiles in ts are about 3tb
[18:11:45] <nosy1> these are stored on an nfs volume
[18:11:52] <Silke_WMDE> oh hi nosy1! cool!
[18:11:53] <nosy1> so speed is not too fast
[18:11:56] <nosy1> hi Silke_WMDE
[18:12:13] <Coren> Hum, I have plenty of space on the hardware slated to receive the Lab's postgresql server.
[18:12:21] <nosy1> it needs to be shared between tile renderer and web servers
[18:12:25] <akosiaris> and how often do they get expired ?
[18:12:45] <akosiaris> mod_tile does that I suppose ?
[18:12:48] <nosy1> do they expire at all?
[18:12:52] <akosiaris> they should be
[18:12:59] <akosiaris> or else what is the point ?
[18:13:07] <nosy1> we have some tiles that are never expired like hillshading
[18:13:09] <nosy1> others
[18:13:10] <Kolossos> No they exoire only if necessary.
[18:13:14] <nosy1> ill check the config
[18:13:38] <scfc_de> So how long does it take to get extra disks up and running?
[18:14:12] <akosiaris> when we get an actual installation not much. but we are not there yet. I just took over the project so...
[18:14:32] <Kolossos> In average I would say the expire all 2 months, but in busy areas shorter.
[18:14:36] <Coren> Also, I can easily earmark 5T or so on the labs NFS server, but that doesn't scale all that well. We probably want to use varnish for this.
[18:15:21] <akosiaris> Coren: my point exactly
[18:15:50] <Silke_WMDE> akosiaris Have you been in touch with Kolossos and also with Kai?
[18:15:52] <scfc_de> I was under the impression that there is already an installation (cf. https://wikitech.wikimedia.org/wiki/OSM_Tileserver#Extended_State_of_Things_-_2013-11-27, "Hardware is already bought and isn't right").
[18:15:55] <akosiaris> but this is all about the OSM production infrastructure. I am kind of confused as to what this has to do with Toolserver
[18:16:04] <nosy1> ok i look at the config for mod_tile
[18:16:08] <nosy1> should i paste?
[18:16:09] <Kolossos> TS statics can be found here: http://munin.toolserver.org/OSM/ptolemy/index.html
[18:16:31] <akosiaris> nosy1: niah. Just send me a link to pastebin at some point
[18:16:39] <nosy1> ok
[18:16:43] <Kolossos> Delivering : 100 Tiles/s Rendering: only 1 metatile /s
[18:16:46] <Coren> akosiaris: Indirectly, lots of TS tools are pending on having Labs replicate the OSM infra.
[18:17:26] <akosiaris> which means those tools will start using the OSM production infra when it is ready ?
[18:17:28] <Coren> (And also, lots of what the prod infrastructure is meant to replace is on the TS)
[18:17:29] <nosy1> akosiaris: http://pastebin.com/rTzhdjRs
[18:17:51] <akosiaris> Coren: ah ok
[18:17:55] <nosy1> we are also delivering extra tiles
[18:18:04] <nosy1> like hillshading etc that dont expire
[18:18:24] <nosy1> but that is simply served via web server off the nfs store
[18:19:39] <tfinc> another blocker that the ops team wisely identified was that in order for us to really support OSM we need to build a team for it
[18:19:51] <tfinc> any by OSM i mean maps stuff in general
[18:20:01] <tfinc> i've started on a proposal to erik for this
[18:20:16] <Silke_WMDE> This >> <Coren> (And also, lots of what the prod infrastructure is meant to replace is on the TS) is one of the reasons this is so important to us - to know that/when those tools will be able to migrate
[18:20:17] <tfinc> is there anyone about that would want to help me with it ?
[18:21:25] <Silke_WMDE> tfinc: Have you presented this idea on maps-l?
[18:21:58] <Silke_WMDE> sounds good -> maybe there are helpers over there.
[18:22:06] <tfinc> Silke_WMDE: nope. its just been brewing in my head right after an irc talk with brandon
[18:22:23] <scfc_de> tfinc: I think the dominant question is time. If WMF starts now to assemble a team to discuss the requirements for a server, when will the whole shebang be actually finished?
[18:22:31] <tfinc> he correctly identified that taking it past the TS migration will require a team
[18:22:35] <nosy1> as i am a noob too i had very constructive help by apmon
[18:22:41] <tfinc> i need help in thinking about what that means resources wise
[18:23:21] <akosiaris> someone then who knows the requirements ?
[18:23:27] <tfinc> scfc_de: i think that this fiscal year it wouldn't be any more then a contractor. next fiscal i would look at a full time team pending approval of course
[18:23:41] <tfinc> id love any leads to anyone within OSM who would be up for this
[18:23:49] <Kolossos> If TS migration is sucessfol , I'm surely there to help to build an OSM team.
[18:23:57] <tfinc> thanks Kolossos
[18:24:23] <tfinc> but none of my search is blocking the ops hardware bits. its just to give us extra hands when we need them
[18:24:51] <Silke_WMDE> scfc_de is right -> the time issue. I want to know as soon as possible if we can stick to the roadmap
[18:24:53] <tfinc> and were trying to be conscious of how much time the data center move is taking from the ops team
[18:25:20] <Silke_WMDE> and if not I would like to get that information out there asap, too
[18:26:04] <Silke_WMDE> I feel, it's not in my/our-TS-WMDE-community hands, but it's in yours at WMF
[18:26:08] <tfinc> i'll let akosiaris speak to the ops timeline/availability as he'll know that best
[18:26:32] <tfinc> akosiaris: what can the rest of engineering do to help unblock you guys ?
[18:27:05] <akosiaris> plural is not really needed. I am solo on this seems like
[18:27:13] <Silke_WMDE> akosiaris tfinc And what can the volunteers do to help?
[18:27:26] <tfinc> great point Silke_WMDE
[18:27:31] <akosiaris> ops team does not really have that much of resources right now.
[18:27:48] <akosiaris> my first point is that I would like requirements.
[18:28:05] <Silke_WMDE> ?
[18:28:10] <scfc_de> akosiaris: Requirements = disk space? Network I/O? CPUs?
[18:28:15] <akosiaris> I picked up the project and my only goal up to now is to build an OSM Tileserver
[18:28:22] <Silke_WMDE> I see
[18:28:51] <akosiaris> and that is about it. So any help pointing out what is really needed would be great
[18:28:52] <tfinc> akosiaris: the TS OSM tile servers are pushing roughly 100 tiles/sec. thats all that we need for now
[18:28:59] <ireas> hi there and sorry for being late (had a phone call to make)
[18:29:15] <nosy1> at ts we have one system with postgres and tile rendering. if you could have 1 or 2 db servers and 2 tile renderers that would be a good starting point?
[18:29:16] <Silke_WMDE> hi ireas, we're talking about OSM right now.
[18:29:19] <tfinc> akosiaris: brandon has all the space specs
[18:29:36] <Silke_WMDE> ireas: I'm collecting other questions in a query
[18:29:45] <ireas> hi Silke_WMDE, okay!
[18:30:05] <tfinc> akosiaris: i think any scoping past the initial move of tools + breading room will take too long
[18:30:16] <tfinc> start small and then we can look at team for this next fiscal
[18:30:24] <akosiaris> breathing room being ? 10% ? 50% more ?
[18:30:26] <tfinc> bradon was convinced that the existing hardware minus storage was enough
[18:30:39] <Coren> Then perhaps the solution is a different approach?
[18:30:52] <tfinc> Kolossos: i've seen the current graphs, what's the year over year growth like ?
[18:31:03] <Silke_WMDE> tfinc akosiaris What's the chance for an OSM-experienced volunteer to get (root) access on the tileserver to help?
[18:31:07] <Coren> Right now, we're looking at "lets support OSM in production" which is an important step, but isn't a requirement.
[18:31:30] <Coren> Perhaps we should concentrate on "Let's get TS-equivalent support in Labs for OSM" as a first step?
[18:31:32] <nosy1> it is a requirement i guess
[18:31:57] <Silke_WMDE> +1 for the toolserver equivalent in a first step
[18:32:04] <nosy1> yes
[18:32:06] <scfc_de> Coren: +1. We would also get good performance data from that.
[18:32:26] <tfinc> yup thats why i was saying "i think any scoping past the initial move of tools + breading room will take too long"
[18:32:30] <akosiaris> Silke_WMDE: not that bad. He would have to sign NDA i think, but that is up to legal. Other than that he won't be the first volunteer with root access
[18:32:32] <Coren> Yes, we absolutely /want/ to have OSM be a "real" production supported system.
[18:32:34] <Kolossos> Tfinc: The groth over the last year was negative. But for OSM I believe you need a factor 2 for 2 years in data grow.
[18:32:55] <tfinc> so less cpu wise but more for storage
[18:33:47] <tfinc> anything else wont fit in the timeline and can be done later
[18:33:54] <Silke_WMDE> akosiaris Ok, I'll get back to this root access question via e-mail
[18:33:58] <tfinc> when resources appropriately
[18:34:03] <tfinc> resourced*
[18:34:10] <Coren> So my first question is "What is needed to simply move the current OSM things running on the TS"? I know we need a postgres server (that's already planned anyways) What else?
[18:34:15] <akosiaris> the toolserver timeline that is right ? June 2014 ?
[18:34:36] <Silke_WMDE> akosiaris: yep
[18:34:38] <Silke_WMDE> :D
[18:35:32] <scfc_de> Coren: The PostgreSQL will only be available after the Labs move to eqiad?
[18:35:56] <Silke_WMDE> akosiaris: "June" basically means that it's WMF's call to pull the brake *now* if necessary.
[18:35:59] <Coren> scfc_de: Well, it could be made available now but there seems to be little point to it (and I'd have limited availability until the move)
[18:36:28] <Coren> Silke_WMDE: I don't think that's needed if we restrict the scope to exclude "make OSM a production-level service" for now.
[18:36:48] <scfc_de> Coren: Okay; and we have 3 TByte available on the NFS servers at eqiad?
[18:36:57] <akosiaris> that practically means just to move everything to labs. However that may not be possible
[18:37:10] <akosiaris> I see 20MB/s read only on ptolemy
[18:37:23] <akosiaris> which will probably kill performance on the rest of labs ?
[18:37:26] <Coren> scfc_de: I can put 5T aside for OSM without even touching my planned allocation. I have over 80T of total space to allocate to labs.
[18:37:33] <Silke_WMDE> Is it okay to give this topic 5 more minutes and then come to questions not concerning OSM but other migration issues?
[18:37:34] <akosiaris> IOPS ?
[18:38:30] <Coren> scfc_de: /not/ counting database space.
[18:39:10] <scfc_de> akosiaris raises an important point about I/O. Could Labs handle this? I think eqiad has more compute nodes?
[18:39:49] <Coren> scfc_de: It does; but also both the storage and the new databases live on physical hardware which is nowhere near as I/O bound as the VM infrastructure.
[18:41:14] <scfc_de> Network links can be saturated as well :-).
[18:41:16] <Silke_WMDE> Talking about the datacenter - Pathoschild has this question: [19:13:14] <Pathoschild> Hi Silke. A question for the office hours: what is the status of the datacenter migration? It's preventing some tools from migrating from Toolserver; see https://bugzilla.wikimedia.org/show_bug.cgi?id=55929
[18:41:41] <Silke_WMDE> can you give us an update please?
[18:42:24] <Coren> Silke_WMDE: The short of it: we're currently bottlenecked on a contractor for the actual Openstack setup in equiad, but Tool Labs is the first project that will be moved (more precisely, duplicated) on the new infrastructure as soon as it is live.
[18:42:49] <Coren> Silke_WMDE: I would normally say 3-4 weeks, but we should probably add one or two more to this to be safe.
[18:42:55] <Silke_WMDE> oh, so you haven't moved yet?
[18:43:20] Silke_WMDE was obviously hibernating
[18:43:27] <Coren> Silke_WMDE: Not completely no, lots of the support infrastructure is ready (fileserver, databases) but we are pending on Openstack.
[18:43:41] <Pathoschild> Coren: so current estimate is early March or so?
[18:44:16] <Coren> Pathoschild: That sounds about right. I am still hoping for end February as I had originally planned, but mid-march is more realistic.
[18:44:34] <Pathoschild> Alright, thanks.
[18:44:57] <Coren> Although "early adopters" will probably able to try moving experimentally a little earlier than that (and will help work the kinks out in so doing)
[18:45:02] <Kolossos> Can we end the OSM part with a schedule for the OSM-db in Labs or production as a first step?
[18:45:42] <Silke_WMDE> akosiaris tfinc That's a question for you two
[18:45:53] <Pathoschild> Coren: yep. All my tools are already running on Tool Labs (alongside Toolserver), but the latency prevents normal use right now.
[18:46:07] <tfinc> akosiaris: you'll know this better then me. let me know if i can help
[18:46:52] <akosiaris> on my part, I am supposed to provide spare time on this. So with a definite danger of being wrong, I 'd say at least 3-4 weeks...
[18:47:24] <Silke_WMDE> ... to get into the topic and set up a schedule, right, akosiaris?
[18:48:01] <Kolossos> no i hope to get a running db.
[18:48:09] <Silke_WMDE> ah
[18:48:11] <akosiaris> ahaha.
[18:48:15] <Coren> akosiaris: I have hardware earmarked for the postgres server; if you want to take ownership of this you'll get it all the sooner. :-)
[18:48:15] <akosiaris> the second one
[18:48:39] <akosiaris> Coren: remind me what is postgres server is for aside from OSM
[18:49:13] <tfinc> akosiaris: can the mobile team and/or volunteers help with this ?
[18:49:14] <Coren> akosiaris: It's meant for OSM mostly, though the deployment-prep people are happy to be able to test builds against postgres as well so it's designated as "postgres for labs".
[18:49:40] <akosiaris> the mobile team ?
[18:49:48] <Silke_WMDE> akosiaris: It would be good to get regular status updates in bugs on the ML, even if it's "small news". Just for transparency.
[18:50:01] <akosiaris> Coren: ok thanks
[18:50:17] <Silke_WMDE> Here: ireas has questions:
[18:50:29] <akosiaris> Silke_WMDE: I 'll do my best
[18:50:32] <Silke_WMDE> cross-wiki joins – current status (see https://wikitech.wikimedia.org/wiki/Nova_Resource:Tools/Help#Joins_between_commons_and_wikidata_and_other_project_databases
[18:50:36] <Silke_WMDE> thanks akosiaris
[18:50:39] <scfc_de> To clarify: We are talking about setting up OSM being replicated to the PostgreSQL server in three to four weeks?
[18:50:47] <tfinc> akosiaris: for instance max semenik did a ton of work with faidon on this in the past
[18:51:50] <akosiaris> scfc_de: more like moved ? not replicated ? And I am still not sure the main labs replicated server is the best option.
[18:51:59] <akosiaris> postgres server*
[18:52:14] <akosiaris> I 'd like to take a look at it first
[18:52:20] <Coren> Silke_WMDE: That's not a technical issue blocking, simply a lack of bandwidth on my part. What /would/ help a lot is a prioritized list of blockers; right now what time I have that isn't going towards migration is assigned "who asked me last".
[18:52:51] <Silke_WMDE> Coren: I see
[18:53:04] <Coren> Silke_WMDE: If I had a list of "this is what is blocking the most migrations", I could work down that list systematically.
[18:53:05] <scfc_de> AFAIUI, the PostgreSQL server on Toolserver is a "replication slave" to the OSM.org master (not regular PostgreSQL replication). Kolossos?
[18:53:34] <tfinc> Coren: akosiaris would it help if i kept a list of that for you guys? i don't have it handy but if you need help wrangling these and don't have the time yourself
[18:53:38] <nosy1> true. it gets diff updates from planet.osm.org
[18:53:42] <Silke_WMDE> Coren: OK, I'll find out and get you a list
[18:53:43] <tfinc> s/handy/written
[18:53:57] <Kolossos> Replication in OSM mean to use diff files each minute.
[18:54:11] <Coren> Silke_WMDE: Alternately, just someone willing to play bug wrangler for the tool labs bugzillas would do just as well.
[18:54:38] <Silke_WMDE> yes, some folks have done a good job to get that started over xmas
[18:54:45] <akosiaris> Kolossos: still haven't looked into that part yet. I might bug you at some point.
[18:54:56] <Coren> Silke_WMDE: André does his best, but he can't guess at relative priority vs toolserver.
[18:55:14] <Coren> Silke_WMDE: I saw, that helped a lot. Need moar!!1! :-)
[18:55:21] <Silke_WMDE> ok
[18:55:29] <Silke_WMDE> Coren: ireas again: Java servlets – see https://bugzilla.wikimedia.org/show_bug.cgi?id=54845 (Coren intended to add Tomcat support similar to FastCGI, but I don’t know the current status)
[18:56:08] <Kolossos> akosiaris:You can ask me everything.
[18:56:18] <Coren> Silke_WMDE: I'm running out of resources in pmtpa - our servers are full. The positive side is that tomcat support is not /complicated/, and should be quick to deploy in eqiad.
[18:56:38] <Coren> (Since I can reuse the lighttpd infrastructure almost verbatim)
[18:57:28] <tfinc> heads up that i'll have to step away @ 11:00 to head to the architecture summit
[18:57:56] <Coren> andrewbogott has done a stellar cleanup job that will allow us to tide over until the migration, but we're still short on resources. Labs proved to be much more popular than we anticipated. :-)
[18:59:58] <Silke_WMDE_> ah - I was gone.
[19:00:00] <Silke_WMDE_> rehi
[19:00:25] <giftpflanze> mail addresses/toolserver.org addresses? any decision/eta yet?
[19:00:25] <Coren> Rehi
[19:00:28] <Coren> Silke_WMDE: I'm running out of resources in pmtpa - our servers are full. The positive side is that tomcat support is not /complicated/, and should be quick to deploy in eqiad.
[19:00:29] <Coren> (Since I can reuse the lighttpd infrastructure almost verbatim)
[19:00:31] <Coren> andrewbogott has done a stellar cleanup job that will allow us to tide over until the migration, but we're still short on resources. Labs proved to be much more popular than we anticipated. :-)
[19:00:57] <tfinc> heads up that i'll have to step away @ 11:00 to head to the architecture summit [repost]
[19:01:26] <Silke_WMDE> tfinc: thanks for attending
[19:01:46] <Coren> giftpflanze: That's been okayed. It's either going to arrive at the same time as eqiad migration or, if the latter takes longer than hoped, arrive just before. This just coincides with a cleanup of mail handling for labs in general.
[19:01:54] <nosy1> mail addresses can easily be given to the ops. it labs hosts the ts domain it should be doable as well as keeping the htaccess redirects
[19:02:14] <Silke_WMDE> Coren: That sounds a bit like if whole setup is pending until you've moved. :(
[19:02:34] <tfinc> Silke_WMDE: just so that everyone knows that i'm walking out with 2x action item, draft OSM team proposal and float it on maps-l, pass migration list to maps-l to make sure we haven't forgotten anything
[19:02:37] <Coren> Silke_WMDE: Much of it is, annoyingly enough, which makes the current bottleneck all the more frustrating.
[19:02:48] <tfinc> does anyone need any other bits from me ?
[19:02:54] <tfinc> especially wrangling
[19:02:58] <nosy1> and we need a postgres db :D
[19:03:11] <giftpflanze> Coren: will old addresses be kept?
[19:03:42] <Silke_WMDE> tfinc: enjoy the summit
[19:03:43] <Coren> giftpflanze: @toolserver.org? That depends mostly on what WMDE does with the domain -- but in the interim nothing should prevent .forwards
[19:04:27] <nosy1> Coren: i personally would favour to hand you the domain over so you can host aliases and redirects
[19:04:38] <Silke_WMDE> If you ask me, I'd like to stick with the idea we discussed earlier and that came up again on toolserver-l -> that WMF gets the domain
[19:04:47] <Silke_WMDE> :)
[19:05:27] <Coren> Nevertheless, there will be a period until final shutdown where that still points at the toolserver so there will be need for some support there.
[19:06:04] <valhallasw> is there any news on email? this might have been mentioned already, but I was late to the party
[19:06:12] <valhallasw> as in: email on @tools.wmflabs.org
[19:06:33] <Coren> valhallasw: It has. Should be really early March.
[19:06:43] <Coren> ready*
[19:06:47] <valhallasw> OK
[19:07:10] <Coren> We got all the all clears (there were difficult issues about rogue tools putting the WMF's AS on blacklist that needed to be addressed)
[19:07:35] <nosy1> Coren: from my point support for mail etc is not a problem for the final shutdown phase as long as we get the tools migrated i am happy
[19:07:54] <Silke_WMDE> dito
[19:08:31] <Coren> I've got mail sorted as "major irritant" and not "blocker". :-)
[19:08:41] <Coren> (UTRS being special in that regard)
[19:08:58] <nosy1> UTRS?
[19:09:13] <Coren> nosy1: https://en.wikipedia.org/wiki/Wikipedia:Unblock_Ticket_Request_System
[19:09:16] <Silke_WMDE> I'll have to run in a few minutes. Is there anything concrete you want me to do apart from prioritizing in BZ?
[19:09:55] <Coren> Silke_WMDE: I think right now we're stuck mostly on externalities, but the prioritizing will make it so that things get done in the most useful order when things unclog.
[19:10:12] <Silke_WMDE> alright
[19:10:15] <giftpflanze> Silke_WMDE: decision for ts domain?
[19:10:46] <Silke_WMDE> giftpflanze: ^^ If you ask me, I'd like to stick with the idea we discussed earlier and that came up again on toolserver-l -> that WMF gets the domain
[19:11:00] <giftpflanze> *sigh*
[19:11:20] <giftpflanze> that's not a definitive answer
[19:11:29] <nosy1> the most reasonable approach i think to keep mail and redirects even if ts is gone
[19:11:31] <scfc_de> I'd prefer WMDE kept it and let the forwarder run in the EU/CH, for example by the current ISP.
[19:11:49] <nosy1> why?
[19:11:50] <Silke_WMDE> giftpflanze: ok: WMF gets the domain
[19:11:59] <giftpflanze> do you personally have the power to decide that?
[19:12:48] <nosy1> wmde is the owner of the domain
[19:13:11] <Silke_WMDE> giftpflanze: did you read the thread about it on toolserver-l? no one objected clearly. so i feel yes i do
[19:13:25] <giftpflanze> ah, ok
[19:13:37] <nosy1> still i dont see any difficulties in hosting aliases and redirects somewhere else but why would this be important?
[19:14:13] <Silke_WMDE> I have to leave you
[19:14:23] <nosy1> bye Silke_WMDE
[19:14:26] <Silke_WMDE> Thanks a lot for coming
[19:14:40] <Silke_WMDE> and for your questions and answers.
[19:14:41] <akosiaris> same here. If there isn't anything else you want to talk about OSM i am gonna leave too
[19:14:55] <Silke_WMDE> akosiaris: LetÄs stay in touch via mail!
[19:15:04] <nosy1> akosiaris: if you need any specs of the current ts let me know
[19:15:15] <akosiaris> Silke_WMDE: sure
[19:15:20] <akosiaris> nosy1: ok will do. thanks
[19:15:24] <ireas> thanks Silke_WMDE, Coren, nosyl et. al. :)
[19:15:52] <Coren> Thank you all.
[19:15:55] <Silke_WMDE> np. ireas thanks for reminding me ;)
[19:16:23] <Silke_WMDE> see you all Coren akosiaris nosy1 ireas Pathoschild giftpflanze *wave*
[19:16:32] <Pathoschild> Goodbye. :)
[19:16:40] <Kolossos> bye.
[19:16:49] <Silke_WMDE> oh ... and Kolossos! :)
[19:16:49] <scfc_de> nosy1: ATM there is an expectation that toolserver.org mail is not subject to (direct) US interception. Moving the forwarder to the US might cause a lot of drama. And contracting Manuel to host the DNS and set up a mail forwarder will not cost WMDE that much. I don't object to the redirects being hosted by WMF.
[19:17:32] <Coren> scfc_de: They'd still end up forwarded through the Labs.
[19:18:07] <nosy1> scfc_de: they are forwarded to mail addresses in the us in parts?
[19:18:19] <nosy1> scfc_de: what would be your main concern?
[19:18:47] <scfc_de> Only if @toolserver.org is forwarded to @wmflabs.org; if someone sets up a forward to @super-privacy.de, it doesn't cross the pond.
[19:19:20] <scfc_de> nosy1: I'm not concerned about *my* @toolserver.org mail (did never use it), only the potential drama of others :-).
[19:21:20] <nosy1> scfc_de: i never thought that other mail was going to these addresses then tool problems
[19:21:25] <nosy1> do you see a problem there?
[19:22:08] <nosy1> and why should people not use pgp?
[19:23:06] <giftpflanze> Coren: will labs addresses be also per user or only
[19:23:14] <giftpflanze> per tool?
[19:23:51] <Coren> giftpflanze: Both, but fixed to go to the wikitech account's email address (so no arbitrary forwarding)
[19:24:09] <Coren> giftpflanze: With mail to tools going to its maintainers.
[19:25:28] <giftpflanze> Coren: so, to all maintainers without flexibility?
[19:26:50] <Coren> giftpflanze: I'm looking at a way to set a forward address for tools, provided it's verified.
[19:26:59] <Coren> giftpflanze: But first iteration will be to the maintainers.
[19:27:08] <giftpflanze> mhm
[19:27:49] <Coren> Not ideal, but at least it (a) works and (b) is not insane as a default.
[19:28:10] <giftpflanze> hehe :)
[19:30:02] <scfc_de> nosy1: I would assume that (and use a @toolserver.org mail in that way) as well, but some Toolserver users have been very secretive about their doings, so I'm not sure. To be clear: This isn't the most important thing to me, and if mail forwarding would move to the US, the earth would keep circling the sun :-).
[19:30:34] <nosy1> true *rotfl*
[19:31:17] <scfc_de> Coren: Don't we already have forward addresses for tools?
[19:31:26] <scfc_de> /data/project/*/.forward?
[19:31:29] <nosy1> scfc_de: but it might be worth a headsup to everyones personal mail account so people can decide to change the address to some not so sensitive one
[19:32:28] <scfc_de> nosy1: Yeah. The present setup will continue for at least five months, I believe, so there's no rush.
[19:32:59] <Coren> scfc_de: Yes, but that will have to change as part of the requirement to allow incoming email to prevent amplification attacks.
[19:33:12] <Coren> scfc_de: It wasn't an issue while only email generated within labs could escape.
[19:35:00] <scfc_de> Coren: Ah, okay.
[19:38:36] <nosy1> ok if you dont have anything for me ill run too?
[20:12:07] <Pathoschild> Who is adding the log for today's office hours on Meta?
[20:18:13] <scfc_de> Pathoschild: If in doubt, always the on who asks the question :-).
[20:18:27] <Pathoschild> Fair enough. :)