Monday, 2018-08-06

*** EmilienM has quit IRC02:39
*** EmilienM has joined #openstack-tc02:40
*** alex_xu has joined #openstack-tc02:43
*** ricolin has joined #openstack-tc03:54
*** tosky has joined #openstack-tc07:39
*** dtantsur|afk is now known as dtantsur07:59
*** cdent has joined #openstack-tc08:00
*** jaosorior has joined #openstack-tc08:21
*** cdent has quit IRC08:45
ttxAgree, low activity is ok... but to remain included in the release (and as an official project team) those need a human to care enough to push a couple of changes per cycle and make it part of the release. Otherwise it's likely to cause more problems (critical bugs, security issues) that would taint the rest of the official release09:42
*** cdent has joined #openstack-tc10:06
*** ricolin has quit IRC10:31
cdentdims do you already have the health review of kuryr in hand, or is there more to do?10:48
cdentpabelanger: same for you and masakari?10:50
*** cdent has quit IRC11:09
*** cdent has joined #openstack-tc11:23
cdentMy efforts to reach ian bicking have so far failed11:28
*** leakypipes is now known as jaypipes11:28
openstackgerritChandan Kumar proposed openstack/governance master: Add tempest-ansible project under tripleo
*** cdent has quit IRC12:02
*** rosmaita has joined #openstack-tc12:03
*** e0ne has joined #openstack-tc12:08
*** e0ne has quit IRC12:13
*** cdent has joined #openstack-tc12:23
dhellmanncdent : I wonder if someone else has already taken over paste/pastedeploy12:25
dhellmannor forked, I guess12:26
cdentdhellmann: I've been lightly searching for evidence but not yet found any12:26
smcginnisAre there major unaddressed bugs with paste that we are aware of?12:30
cdentsmcginnis: more like expected bugs:
smcginnisI bet they have URLS that don't use https too. :)12:34
dimscdent : havent' reached out to dmellado for kuryr yet. will do that today12:35
smcginnisMost of those paste PRs do look like good things to merge.12:36
*** dtantsur is now known as dtantsur|brb12:36
*** cdent has quit IRC12:39
*** cdent has joined #openstack-tc12:43
*** rosmaita has quit IRC13:10
*** edmondsw has joined #openstack-tc13:13
*** jroll has quit IRC13:19
*** jroll has joined #openstack-tc13:19
*** rosmaita has joined #openstack-tc13:23
*** mriedem has joined #openstack-tc13:41
*** lbragstad has joined #openstack-tc13:44
*** clarkb has joined #openstack-tc13:56
*** spotz has joined #openstack-tc14:08
scaslast week, bringing up the rear, chef openstack branched its stable/queens. publisher issues have precluded an announcement, pending a state of the kitchen14:16
openstackgerritMerged openstack/governance master: I18n Extra-ATCs for Rocky
openstackgerritMerged openstack/governance master: Remove Security project team
openstackgerritMerged openstack/governance master: fix markup for links
openstackgerritMerged openstack/governance master: remove placeholder file from stein goals
*** annabelleB has joined #openstack-tc14:24
mriedemis FFU a SIG now?14:25
smcginnisThere is an upgrades sig I believe.14:25
*** dtantsur|brb has quit IRC14:26
*** dtantsur has joined #openstack-tc14:28
*** e0ne has joined #openstack-tc14:43
*** cdent has quit IRC14:45
*** chandankumar has joined #openstack-tc15:04
*** e0ne has quit IRC15:14
*** cdent has joined #openstack-tc15:17
*** ricolin has joined #openstack-tc15:38
*** jaosorior has quit IRC15:48
*** jaosorior has joined #openstack-tc15:49
*** dklyle has joined #openstack-tc15:57
fungiinteresting... cyborg doesn't do weekly meetings on irc as far as i can find from eavesdrop (which is fine if they don't need any), but their drivers subteam meets on zoom?
smcginnisI think I saw that. I haven't looked lately, but I think most of the cyborg team are/were from China.16:28
smcginnisThat probably results in less IRC activity for the project.16:28
ttxmriedem: yes FFU is part of the "Upgrades" SIG, and there was a recent thread on the ML about the SIG recent [in]activity16:31
mriedemwait wait wait, you're telling me creating a sig didn't magically make people appear to work on stuff?!16:32
mriedemsorry, i couldn't help16:33
mriedemfungi: smcginnis: between lenovo and hauwei working on cyborg in china, i wouldn't be surprised if most discussion happens over wecaht16:33
fungittx: different from the thread on the self-healing sig [in]activity?16:35
ttxmriedem: that thread develops an interesting theory on why SIGs are popular at events but less popular in between16:37
ttxmriedem: that does not make them necessarily useless16:37
mriedemi couldn't easily find a thread on upgrade sig16:37
ttxfungi: I think so yes16:37
mriedemand inactivity16:37
* ttx digs16:37
ttxI may be confusing two threads16:37
smcginnisI thought I saw something just within the last week or two.16:38
mriedemthings are popular at events because you have like-minded people in a room talking about solving a shared problem,16:38
mriedembut then everyone goes back to reality with no time to work on said things,16:38
mriedemis that the gist?16:38
ttxThat would be the upgrades SIG thread ^16:39
mriedemoh heh i saw that subject but assumed it was a literal reboot/poweroff16:39
mriedemas in nova reboot :)16:39
ttxtopic overload > reboot16:39
ttxwhich is where the theory is discussed16:40
ttxmriedem: yes, that's the gist of it16:41
ttxLooks like temp just dropped below 100F/38C and I can try to go outside again16:42
smcginnisSounds crazy there. I've only been in the fall and spring when it's still a bit chilly, so can't even imagine it being that warm there.16:43
*** e0ne has joined #openstack-tc16:47
*** evrardjp has quit IRC16:51
*** diablo_rojo has joined #openstack-tc17:06
*** diablo_rojo has quit IRC17:17
*** diablo_rojo has joined #openstack-tc17:23
dhellmannEmilienM : did anyone from the murano team ever respond to your health check email from 17 july?17:27
EmilienMdhellmann: nope17:28
*** annabelleB has quit IRC17:28
dhellmannEmilienM : I found something from Rong Zhu talking about IRC meetings and ZTE. I guess that was it?17:29
EmilienMdhellmann: sorry, yes17:30
EmilienMdhellmann: I got no answer from Tricircle though17:30
EmilienMdhellmann: I'll work on that this week17:32
dhellmannEmilienM : ack17:33
dhellmannttx: did you have a chance to contact storlets? I can do it if you're still looking for ways to cool off17:33
*** annabelleB has joined #openstack-tc17:40
*** evrardjp has joined #openstack-tc17:41
dhellmannsmcginnis , cdent : the detail in is coming out ok. I would like to have enough there to share it with anyone we were going to try to invite to participate in a discussion, so they can see the sorts of questions that would be raised and prepare in advance -- we're all likely to find the discussion more productive that way. Do you have more to add to the document?17:48
cdentdhellmann: I've added one thing17:51
cdentI'll keep thinking17:52
*** dtantsur is now known as dtantsur|afk17:53
dhellmannok, thanks17:56
dhellmannthe hand signals bit from seems interesting; I wonder how that would work out for us18:00
*** e0ne has quit IRC18:05
cdentdhellmann: those are useful, but the critical factor is having an active facillitator18:05
cdent(one who has the power to "pause")18:05
cdenthaving the signals would help to lend some authenticity to the idea of facillitation18:06
*** ricolin has quit IRC18:11
*** e0ne has joined #openstack-tc18:14
mriedemdtroyer: mnaser: fyi - i plan on charting up some of that for high level takeaways, and then i'll post links to those in the ML18:43
mnasermriedem: that's really interesting, thank you for taking the time to building all that out, some really interesting patches18:45
mnaserdtroyer: will the starlingx team be at the ptg by any chance18:46
mriedemdean will be :)18:46
smcginnisdhellmann: I will add more to the etherpad if/when I can think of more. I know there is more to add, but haven't been able to free up enough brain cycles for it.18:48
dhellmannsmcginnis : ack, thanks18:49
mriedemsmcginnis: coincidentally starlingx has a feature to allow swapping out brain cache18:49
dtroyermnaser: we will be there with a full day room on Wed18:51
dtroyermriedem: stx also cures the common boredom and re-programs your LED lights to flash the status of your edge cloud nodes18:53
mriedemwe have gone cyan18:53
mriedemnotify the canucks18:54
dtroyermriedem: that spreadsheet… you did that from the squash?18:55
dtroyermind if I put some mysterious numbers in column F?  I want to start matching that up with the patch files I'm still trying to get cleared to publish18:58
mriedemdtroyer: yes18:59
mriedemstarted going through the diff from top to bottom and clone the repo locally for when i needed context on linking changes from the api down the stack18:59
smcginnisWe should probably talk about py37 at some point.18:59
mriedemdtroyer: you should have comment auth on the spreadsheet, is that good enough?19:00
dtroyermriedem: yes.  we have a master sheet with storyboard links and review pointers (for those that exist) that I want to tie this all together19:01
dtroyertried to do it directly in storyboard but it isn't quite up to it without serios query scripting, PM19:01
cdentsmcginnis: yes19:02
dtroyers not liking that route19:02
*** annabelleB has quit IRC19:29
*** annabelleB has joined #openstack-tc19:34
*** zaneb has quit IRC19:37
fungidtroyer: i'm curious about the use case for the server group messaging feature. was there some reason traditional networking wasn't possible?19:41
fungiallowing servers to pass messages back and forth sounds like the quintessential description of traditional computer networks19:42
fungiso i have to assume there's some corner case where that simply wasn't possible, otherwise that's a rather massive wheel to be reinventing from scratch19:42
* fungi is reminded of a time long, long ago when he had a data center full of filtering ethernet bridges with control channels over rs232 serial, but that was entirely due to hardware limitations19:45
dtroyerfungi: context?  I have too many things in my head atm…19:45
fungidtroyer: row 24 in mriedem's spreadsheet19:45
fungi"Server group messaging. Creates a channel between the guests and their host in a server group so that guests can send messages to each other. Compute would cast the messages to conductor which would then broadcast the message to all servers in the group."19:45
dtroyeralso, The People Who Know are supposedly reading the starlingx-discuss list these days…19:45
mriedemcfriesen is out this week19:46
mriedembut he knew about it19:46
dtroyerI'm trying to match those lines up to the specific patches...19:46
mriedemsomeone needed a low-latency way for servers to communicate with each other19:46
mriedemso they create a channel between the compute host and the guests,19:46
mriedemand route the messages through nova-conductor using rpc19:46
dtroyerthey also cranked down wait times to 5ms in a number of cases, so that isn't surprising19:46
mriedemcfriesen said it was more of a pain in the ass than it was worth,19:47
fungiforwarding messages over rpc through the conductor was lower latency than ethernet?19:47
mriedemand they were likely going to deprecate/remove it19:47
mriedemfungi: nova-conductor was really just the orchestrator,19:48
mriedemnova-compute would get the message from the guest,19:48
mriedemcast to conductor to find the other members of the group to route to them19:48
fungiahh, so maybe they had faster networking between the compute nodes than they allowed the guests to get interfaces directly in19:48
mriedemfirst slide i'm doing is a disclaimer to say "i'm not authority on this"19:49
fungi"direct questions to the starlingx-discuss ml" ;)19:50
jrolledge sites tend to have limited bandwidth, too. so could be that letting nova do the fan-out took up less bandwidth19:50
*** zaneb has joined #openstack-tc19:51
jrolllike, each edge site would only get the message via WAN once, instead of a guest in site A spamming N messages to site B, where site B has N guests19:51
mriedemheh i didn't know existed19:51
mriedemnor do i plan on subscribing19:51
jrollyou sure seem interested in discussing it :P19:51
mriedemafter spending ~5 days looking at this, i need to feel justified in the time spent19:52
fungientirely fair19:52
mriedem0 responses19:52
dtroyerheh, pm's don't always start the _right_ discussions with 3 people subscribed to a list :)19:53
mriedem"question: WHO IS READY TO ROCK AND/OR ROLL?!?!!"19:53
* fungi feels like he's just stepped into a böc concert19:54
* dtroyer is ready to try to open-source Tickemaster's seat selection algorithms next…19:54
clarkbjroll: multicast :P19:56
jrollmulticast across the internet, I like it :P19:56
clarkbI seem to recall we had to dosimilar for an ISS experiment at PSU once upon a time19:57
clarkbacross the internet and space!19:57
jrollheh, sounds fun19:58
cdentI miss the mbone19:58
dtroyerISS == extreme edge19:58
jrollthe edgiest of edges19:58
mriedemdtroyer: btw, since you're not in the nova channel but your name is tied to the vif_model thing20:01
mriedemnewton: and ocata:
dtroyermriedem: I owe you $COLD_BEVERAGE or three for this spreadsheet…20:01
dtroyermriedem: ah, I should join there, that was meant to be a place-holder for WRS to fill in20:02
zanebdhellmann: any meeting ground rules that involve pointing at people are inappropriate in a multicultural environment20:03
fungihuh... so the next board meeting happens during tuesday of the ptg? guess i won't be dialling in to listen to that one:
fungino, wait, that's the tuesday _after_ the ptg20:07
fungiso i guess i can. excellent20:07
clarkbya the week after20:10
*** e0ne has quit IRC20:11
dhellmannzaneb : good point. I was more thinking about the general idea than the specific gestures, but I wouldn't have thought of that issue20:15
*** annabelleB has quit IRC20:21
*** annabelleB has joined #openstack-tc20:24
cmurphypabelanger: mnaser I added some notes to
*** edmondsw has quit IRC21:15
*** harlowja has joined #openstack-tc21:21
*** cdent has quit IRC21:40
fungiTheJulia: i caught up with refstack people in their irc channel and updated with a link to the discussion and a summary of the present state of thinking there21:45
TheJuliafungi: thanks, it completely slipped my mind :(21:52
mriedemdtroyer: mnaser: bombs away22:03
*** annabelleB has quit IRC22:07
* dtroyer ducks, and coincidently just finished the first pass through that spreadsheet...22:10
clarkbmriedem: dtroyer an interesting one to me is the ionice for image conversions. Might be better to avoid conversions entirely and use the appropriate image format for nova?22:19
clarkbthis is something we've run into with infra as well and for our uses specifically we've found that giving the cloud the format it wants its typically the best approach22:20
dtroyermriedem: nice summary, especially from a squashed pile-o-bits.  I'm still pushing to publish the individual patches, with interesting (useful) commit messages…22:20
dtroyerclarkb: soooo many things in TC were done directly for making customers happy without regard to such thinking…22:21
dtroyeroops, I said that out loud22:21
* dtroyer grasps for a bit-eeraser22:21
mriedemclarkb: heh, i threw that in at the very end just b/c i had the space22:30
mriedemdidn't think ionice would be that interesting22:30
clarkbmriedem: I mostly just mention it because maybe its a user education thing (I know it took us a while to figure out, why do our node launches all fail once a day? its because we uploaded qcow2 and cloud needed raw so then it took 15 minute sto convert the image and our builds timed out)22:30
clarkbon a nova deployment side the default is to launch raw, but we don't really say that to anyone that raw is the thing that nova expects by default22:31
mriedemwhere does the cloud tell you it needs raw?22:31
clarkbmriedem: it doesn't thats the other problem22:32
mriedemwell there you go22:32
clarkbmriedem: we've learned through trial and error and encoded it into openstack client config22:32
clarkbsome clouds will tell you if you ask nicely22:32
fungimostly just through meat-based apis22:48
*** tosky has quit IRC23:02
*** annabelleB has joined #openstack-tc23:07
mriedemi don't have great ideas on how to expose that to non-admin users23:31
mriedemwould there be security exposures to doing so?23:31
mriedemalso, isn't the format in the image meta? and couldn't an admin create host aggregates for images of a specific format that should go to hosts configured for those formats?23:32
mriedemno one probably does because they don't normally timeout?23:32
clarkbya once the image conversion is cached its fine23:33
clarkbits just during that short period of time it gets sad23:33
*** annabelleB has quit IRC23:56

Generated by 2.15.3 by Marius Gedminas - find it at!