--- Log opened Wed Dec 10 11:05:57 2008
-!- wwoods changed the topic of #fedora-meeting to: Fedora QA meeting | init11:06
-!- mether [n=sundaram@] has joined #fedora-meeting11:06
-!- openpercept [n=openperc@unaffiliated/openpercept] has joined #fedora-meeting11:06
wwoodsf13, jds2001, jlaska, poelcat: ping11:06
wwoodsanyone else who's around for the QA meeting say "hello" for the logs11:06
* f13 11:06
* jds2001 11:06
jwbwwoods, i have a quick question for QA11:07
wwoodsjwb: what's that?11:07
* viking_ice_ here11:08
wwoodsis the list you want11:08
jwbcan it be removed?11:08
wwoodssure, if you know how11:08
-!- mbonnet [n=mikeb@c-76-19-115-225.hsd1.ma.comcast.net] has quit [Read error: 60 (Operation timed out)]11:08
* jwb goes to file infrastructure ticket11:08
-!- mbonnet [n=mikeb@c-76-19-115-225.hsd1.ma.comcast.net] has joined #fedora-meeting11:09
-!- spoleeba [n=one@fedora/Jef] has quit ["Leaving"]11:09
wwoodsit turns out to be really unhelpful to separate "testers" from "qa"11:09
wwoodswhich should have been obvious in the first place11:09
wwoodsbut oh well, lesson learned11:09
-!- Gaaruto [n=Gaaruto@fedora/Gaaruto] has quit [Remote closed the connection]11:09
wwoodsokay so. agenda is here: https://www.redhat.com/archives/fedora-test-list/2008-December/msg00297.html11:10
viking_ice_hehe first post in a year..11:10
wwoods1st item: review CommonBugs11:10
-!- wwoods changed the topic of #fedora-meeting to: Fedora QA meeting | https://fedoraproject.org/wiki/Common_F10_bugs11:11
wwoodslook over the list; is anything missing? are any of these fixed?11:12
f13I wonder if it's worth temporarily putting up the broken PackageKit and broken deps bugs11:13
wwoodsyeah, might be worthy11:13
wwoodsanyone have a good summary of the problem / link to bug report?11:13
jds2001and the dbus stuff, but I think that's fixed already, right?11:13
-!- leitz [n=leitz@adsl-068-209-163-224.sip.clt.bellsouth.net] has joined #fedora-meeting11:14
f13jds2001: provided the last updates push didn't result in broken deps11:14
f13the jury is still out, I'm told the last push did "weird things"11:14
wwoodsthe last push itself, or the last-pushed package did?11:15
-!- Gaaruto [n=Gaaruto@fedora/Gaaruto] has joined #fedora-meeting11:15
wwoodsnobody? I know nothing about this bug11:18
f13the last push itself11:18
f13which finished sometime yesterday evening11:18
wwoodsf13: ah, that's less-bad then11:18
f13but still don't know if all is well11:18
wwoodscan you summarize the problem? or point me at a bugzilla bug?11:19
f13man there are a bunch.11:20
f13actually walters would be the best person for that11:20
jds2001http://tinyurl.com/dbusbug http://tinyurl.com/dbusbluez11:20
jds2001that's what's in the topic of #fedora right now.11:20
nirikdbus security update, breaks all dbus using apps including PackageKit.11:20
wwoods*all* of them? zang11:21
niriknew PK pushed out, but still other apps have issues (although many of them are fixed now, but it's unclear)11:21
ivazquezAnd the fixes are in varying states of "pushed".11:21
wwoodsI thought it just bonked PK11:21
nirikalso some system-config* and a report of NM, and...11:21
wwoodsjds2001: do you want to summarize for CommonBugs or should I ask walters11:22
-!- JSchmitt [n=s4504kr@fedora/JSchmitt] has quit [Remote closed the connection]11:22
-!- nphilipp [n=nils@] has quit [Read error: 145 (Connection timed out)]11:22
jds2001wwoods: i would ask walters since I really don't feel qualified to summarize that.11:23
wwoodsI'll ask him11:25
wwoodsand, for the record: no amount of magical QA sauce would have helped here, since the developer(s) decided to skip the entire QA process11:26
wwoodswalters knows this, no need to hurl further blame11:26
viking_ice_So the flaw lies in the process..11:26
wwoodsthe QA process is supposed to be *technically* optional but you're not supposed to skip it unless you have a really damn good reason11:27
* nirik notes that he is the one who approved it from security... but even if I had tested it I wouldn't have seen the issue, as I don't use packagekit and seldom use system-config-* 11:27
wwoodsand you're 100% sure you're not going to break anything11:27
f13the QA blame, or maybe the infrastructure / releng blame, does come into play with the resulting broken deps of PackageKit and gnome-apckagekit11:27
wwoodsif you skip the process, you accept responsibility for the outcome11:28
f13I talked with the board about this yesterday because the issue came up11:28
wwoodsbut yes, some amount of automated checking of deps would still be great11:28
f13and I'm charged with getting something in place this release cycle11:28
viking_ice_no but it points if further restrictions are perhaps necessary..11:28
f13so I'm going to be working with lmacken, skvidal, the QA group, and well myself to get /something/ in place to help developers from shooting themselves in the foot with broken deps11:29
f13viking_ice_: I really don't think more hurdles and red-tape is really the answer.11:29
jds2001me either.  This can be solved technically.11:29
wwoodsno, the real upshot of this problem is: 1) it is very damn clean to everyone that you should NOT skip the updates-testing phase11:30
wwoodsand 2) we really do need some kind of automated dep checking in bodhi etc.11:30
wwoodsboth of which we knew beforehand11:31
nirikso can security updates go to testing? I know at one point they automagically went to stable. We should confirm that they can go to testing at least.11:32
viking_ice_Could we not make every package to have a time in updates-testing just some minimum?11:32
f13I don't think so11:32
viking_ice_Sec updates stay for 6 perhaps 12 hours..11:32
wwoodssecurity updates generally need to go out as quick as possible, which is why we allow people to skip the testing phase *if* they are *really* sure they know what they are doing11:32
f13(to forcing things into -testing)11:32
wwoodsgenerally the security team knows what they're doing11:32
f13I think at this point it's the submitters who decide where it goes11:33
f13not the security team11:33
ivazquezI don't think this was decided by the security team.11:33
wwoodsso forcing them to go through extra work for the 98% of the time when they get everything right11:33
f13the security team just verifies that the security info is accurate11:33
wwoodsjust to avoid the 2% of the time when they get it wrong11:33
wwoodsis not useful11:33
wwoodsit just slows down our security updates11:33
nirikyeah, it's submitter and/or bodhi that determines that, not security team11:34
viking_ice_slight delay on sec updates vs potential broken system11:34
wwoodsfalse-positive paradox. it means we slow down all good updates unnecessarily, and doesn't even guarantee that we'll catch the bad ones.11:34
wwoodsit's not useful.11:34
wwoodsit's more useful to provide better tools for testing11:35
-!- abadger19991 is now known as abadger199911:35
wwoodsso the security team (et. al.) can be *more* sure that the update is OK, and test stuff *faster*11:35
wwoodsrather than slowing everything down11:35
wwoodsthe answer is better tools and clearer docs for testing, not forced delays on all packages11:37
wwoodsanyway. ACTION: wwoods will talk to walters about the dbus bug and update Common_F10_Bugs appropriately11:38
wwoodsany further discussion on that?11:38
* nirik would like to have a fudcon session about security and updates...if the approval is usefull and what exactly should be checked, but thats possibly another topic. 11:38
-!- Kick__ [n=karsten@p54A3C740.dip.t-dialin.net] has quit [Connection timed out]11:38
-!- Kick__ [n=karsten@p54A3C740.dip.t-dialin.net] has joined #fedora-meeting11:39
-!- jsmith [n=jsmith@asterisk/training-and-documentation-guru/jsmith] has joined #fedora-meeting11:39
wwoodsnirik: definitely worth discussing at FUDCon - viking_ice_ and I have been talking a lot about getting test plans/scripts into the wiki and/or package CVS11:39
wwoodsbut that's a big topic and we'll cover it some other time11:39
wwoodsnext agenda item, then11:40
-!- wwoods changed the topic of #fedora-meeting to: Fedora QA | automated installer testing11:40
wwoodsyes, once again, automated installer testing.11:40
wwoodsHere's what we've been working with: https://fedorahosted.org/snake/11:40
wwoodsthe goal, for a long time now, has been: automated nightly smoke tests of rawhide11:41
-!- stickster is now known as stickster_afk11:41
-!- hanthana [n=hanthana@] has quit [Read error: 113 (No route to host)]11:41
wwoodsthis is going to require a special lab setup (which I'm putting together here)11:42
wwoodshardware's the easy part11:42
wwoodsthe hard part is getting the software to do what you want11:42
jds2001OK, why not in PHX? (just curious)11:42
wwoodsit could be in PHX but that's irrelevant11:42
wwoodshardware is easy11:42
wwoodsand I dunno if we have space in PHX11:43
f13well, hardware outside of PHX is easy (:11:43
* leitz wonders what PHX is.11:43
viking_ice_Link to PHX please..11:43
f13wwoods: we're supposed to talk to mmcgrath about QA resources in our PHX colo at FUDCon11:43
wwoods(PHX is our colocation facility in Phoenix)11:43
leitzAh, Pheonix.11:43
f13PHX is where koji lives, and where all the Fedora packages live11:43
wwoodsPHX = "public Fedora infrastructure"11:43
f13and it's where the rawhide composes happen, so it's a logical place for the rawhide testing to happen.11:44
wwoodsbut the point of snake is to have something that you can run anywhere11:44
wwoodsthis should work so long as you have power control - either a remote power adapter or, you know, sitting next to the machine and flipping the switch11:44
leitzMight be better to not have it in PHX so there's less chance of something escaping and busting production.11:44
wwoodsand a way to restore a dead machine - either PXE-booting an automatic install or just installing from a LiveCD11:45
f13leitz: it can be isolated.11:45
wwoodsthe point is that this will run *somewhere* - and it doesn't matter where - and that will generate logs11:45
wwoodsthat people can inspect11:45
wwoodsand it will color some boxes green or red11:45
f13in fact, it should be ran in as many places as possible (:11:45
wwoodsand that's how we'll know if rawhide is broken11:45
jds2001well, we'll know it installs and boot11:46
wwoodsso for that specific definition of "broken"11:46
jds2001whether or not it does anything useful is another thing entirely :)11:46
wwoodsanyone who knows python or has some spare machines and disk space11:46
wwoodsand wants to work on automatic install stuff11:47
* jds2001 has the latter and a little of the former11:47
wwoodsshould definitely email me or come to #fedora-qa and we'll talk about trying to set up a snake-server and what needs doing11:47
wwoodsI think that's going to be my primary goal between now and, say, F11Alpha11:48
* jds2001 already has a snake-server, but it's on an F8 machine :P11:48
f13given that I have a local mirror, cobbler setup, and the ability to manually control power, my office is a good candidate for doing such tests11:48
f13mostly I've been doing the tests in virt11:48
viking_ice_wwoods; cant you create a step by step how to ( thinking materal for testing advanced/expert )11:48
wwoodsAFAIK there are teams inside RH working on some of the other stuff I've been messing with - CrashHandling, test reporting, etc. 11:48
jds2001f13: my "office" is similar, but no cobbler11:49
jds2001manually controlled PXE and all that, though, so cobbler wouldn't be hard11:49
wwoodsviking_ice_: https://fedorahosted.org/snake/wiki/SnakeExamples/ServerSetup etc.11:49
f13ah codeina11:49
wwoodscobbler is kind of weird for QA11:50
viking_ice_wwoods: step by step from a to b11:50
wwoodsit's awesome for datacenters and such11:50
wwoodsbut when you want to install a new distro every day it doesn't really work the way you want11:50
wwoodsbut it's probably the best way to set up the fallback stuff11:51
f13wwoods: how's that?11:51
f13cobbler works great for me to fire up virt guests like crazy11:51
f13new one every day11:51
-!- hadess [n=hadess@cpc1-glfd1-0-0-cust325.glfd.cable.ntl.com] has joined #fedora-meeting11:52
wwoodsa virt guest install of rawhide?11:52
wwoodswell heck11:52
f13using koan11:52
wwoodsmaybe I should reexamine cobbler11:52
f13sudo koan --server localhost --virt --profile=rawhide-http-x86_64 --virt-path /dev/data/rawkvm --virt-name rawkvm11:52
* jds2001 was more using snake to to that11:53
wwoodslast time I looked - and that was more than a year ago - it wasn't well-suited to what we wanted to do11:53
f13the only extra step to do would be: virsh undefine rawkvm11:53
jds2001but even that requires doing a snake-tree update all every day11:53
jlaskayou can use both :)11:53
f13that removes the libvirt entry for "rawkvm" so that it can be re-created the next day11:53
jlaskaI'm trying tou outline how I'm doing it at http://jlaska.livejournal.com/tag/cobbler11:53
f13my cobbler setup has profiles for rawhide-{http,nfs}-{i386,x86_64}11:54
f13and I can add profiles easily based on these that have different package sets, different partitioning, etc...11:54
f13all very very easy to manage11:54
viking_ice_wwoods: if you create some doc ( set up a snake-server or just copy/paste and send me)  discussion  I got my hands full coming up with ext4 testing docs which will end as  "how to test a posix filesystem"11:55
viking_ice_irc discussion11:55
leitzIs it feasible to snake/cobbler from a remote host?11:55
wwoodsf13: yeah, I have to revisit cobbler then11:55
f13leitz: well, my cobbler box is "remote"11:56
wwoodslast I looked at it, it wanted to create a local copy of the distribution and wanted me to recreate the profile daily or something11:56
f13leitz: as in it sits in my office without a head11:56
f13leitz: I ssh into it to launch koan, and then use virt-manager on my laptop to connect to the libvirt on the cobbler box and view the guest vi vnc11:56
f13via vnc11:56
f13wwoods: it grew the ability to have --available-as=nfs://blah11:57
f13or --available-as=http://blah11:57
leitzf13: I'm probably giong to decomission the old servers I have at my feet, so the question is, can I do testing on this from someone else's machines?11:57
-!- kital [n=Joerg_Si@fedora/kital] has joined #fedora-meeting11:57
wwoodswell, hey, I'll be happy if cobbler has grown all the stuff we need to deal with automated installs11:57
wwoodsless work for me11:57
f13wwoods: we should talk at FUDCon11:58
wwoodsokay. ACTION: reinvestigate cobbler as an automated installer tool11:58
wwoodsif that works, we just need remote logging (which we can do a bunch of different ways)11:58
viking_ice_snake or cobbler..11:58
f13oh it has that11:58
jlaskasort of11:58
f13it uses syslog to log everything back to the cobbler host11:58
-!- JSchmitt [n=s4504kr@fedora/JSchmitt] has joined #fedora-meeting11:58
wwoodsjlaska: I assume you've looked at this more recently than I have11:58
jds2001wwoods: he has a whole set of blog entries on it! :)11:59
wwoodsokay I'll do some reading and we'll talk more about this next week11:59
wwoodsand I may get the itch to prototype something new12:00
wwoodsbecause I was never happy about reinventing the wheel in SNAKE12:00
wwoodsugh, gotta get some hardware that'll do KVM12:00
wwoodsstupid BIOS-disabled intel box12:00
viking_ice_jlaska: nice..12:01
wwoodswe'll talk more about cobbler in the coming weeks, I'm sure12:01
jlaskamore to come :)12:01
f13wwoods: doesn't RDU have an intel critter running around?12:01
f13wwoods: who delivers magical sdv boxes?12:02
wwoodsf13: probably, but I've had my fill of SDV wonkiness12:02
wwoodsI want some actual production hardware12:02
wwoodsanyway. reinvestigating cobbler for automated installs. whee!12:03
wwoodsanything else on the topic of automated install testing?12:03
f13not much, other than I need to see if rawhide is working again12:03
viking_ice_first settle on a test system then let's extend..12:04
-!- No5251 [n=No5251@p50990d02.dip0.t-ipconnect.de] has quit [Read error: 110 (Connection timed out)]12:04
wwoodsyeah, last I checked it was a little nasty in rawhideland12:04
f13well I fixed the mkinitrd fallout12:04
wwoodsviking_ice_: right, we'll research the options and pick some way to do the installs ASAP12:04
wwoodsoh, and I still need to talk to the wiki folks about the QA: wiki namespace12:05
f13I just didn't sync yesterday or the day before to see if things started working12:05
f13oh right12:05
f13we got images, we were justmissing squashfs in the kernel, which we shold have today12:05
leitzwwoods: Am I correct in thinking Snake *or* Cobbler? Are the supplementary or does one replace the other for what you need?12:05
wwoodsI had a trac ticket for that once, dunno if it's been closed or if it's just fallow12:05
wwoodsleitz: they're basically the same idea, yeah12:05
wwoodsjlaska and I started writing SNAKE around the same time as Cobbler, and we keep talking to mpdehaan (Cobbler author) about it12:06
wwoodsand we always came to the conclusion that we were doing approximately the same thing, but there wasn't any clear way to merge the projects 12:06
wwoodsor share code12:06
f13one area I think snake has in favor12:06
jlaskaI'm using snake to build the kickstarts still12:06
f13is that there is a snake client that can setup an already installed system when said system isn't on the same PXE network as the cobbler box12:06
-!- fab [n=bellet@monkey.creatis.insa-lyon.fr] has quit ["Leaving"]12:07
wwoodshrm. wonder if you could do that with cobbler12:07
f13it should be easy if you can reach teh cobbler server12:08
f13to query where the vmlinuz and initrd could be found and grabbed12:08
f13maybe this already exists for cobbler and I missed it12:08
f13(I wasn't looking closely)12:08
wwoodsI'll check into that12:08
jlaskaf13: run koan on the client?12:08
f13jlaska: koan is for virt through isn't it?12:08
f13ah, well, there you go12:09
leitzwwoods: Would your time be better spent merging into Cobbler then moving on to the rest of your task load?12:09
wwoodsleitz: I keep thinking that, yeah12:09
wwoodsit wasn't feasible in the past, though12:09
wwoodsmight be time now12:09
wwoodsI liked messing with SNAKE but I'd be happy to have less code to worry about maintaining12:10
wwoodsso I'll bug jlaska for some cobbler/koan help and try to prototype that this week12:10
wwoodsand we'll see how it goes12:10
wwoodswoo, exciting!12:11
wwoodsanything else about automated installs?12:11
wwoodsI think we defined the basic rawhide test goals at the last Boston FUDCon12:12
wwoodsI'll need to see if I can find those again. it's pretty obvious: does it boot?12:12
f13yeah, we just need something to run on the host or in the %post that says pass or fail.12:12
wwoodsbut it's nice to have an explicitly documented set of goals.12:12
wwoodsf13: probably it'll run as an initscript after the system reboots12:12
wwoodsjlaska: aha, thanks12:12
-!- mether [n=sundaram@] has quit [Read error: 104 (Connection reset by peer)]12:12
f13wwoods: yeah, we can easily fetch test scripts in %post of theinstall to be launched upon reboot12:13
wwoodsf13: right, which gives us a platform for any automated testing12:13
wwoodsand then things start getting really interesting.12:14
-!- mether [n=sundaram@] has joined #fedora-meeting12:14
wwoodsanyway. don't want to look too far ahead12:14
wwoodsanything else we need to discuss?12:14
-!- wwoods changed the topic of #fedora-meeting to: Fedora QA | other12:14
-!- mether [n=sundaram@] has quit [Read error: 54 (Connection reset by peer)]12:14
f13this week is when I'll be taking a much closer re-look at our compose processes, to figure out where we can insert some pre-QA12:14
f13I'm going to be eyeballing the added work to verifytree, and things like repoclosure12:15
-!- mether [n=sundaram@] has joined #fedora-meeting12:15
wwoodsrepoclosure? mercy12:15
-!- fbijlsma [n=fbijlsma@h-] has quit ["Leaving"]12:15
viking_ice_One thing Early feature testing..12:15
viking_ice_lets do it :)12:16
wwoodsah yes, early feature testing stuff12:16
wwoodsso all Features need to have a "spec" by Alpha - basically the Scope section needs to say exactly what changes the Feature will make12:16
wwoodsand what it will (and won't) be capable of doing12:17
wwoodsif there's any features that catch your fancy12:17
wwoodshttp://fedoraproject.org/wiki/Features/Dashboard <-- check this page for feature status12:17
leitzSort of like a Use Case.  :)12:17
wwoodsleitz: pretty much, yeah12:17
wwoodsUse Case, Spec, Scope12:18
wwoodswhatever you call it - it needs to be there by Alpha12:18
wwoodsand by Beta, the feature owner (with our help) needs to write a test plan - docs for us testers on how to make sure the feature matches the spec12:18
leitzMy *opinion*, for whatever it's worth, is that the whole bhodi voting thing get tossed.12:18
viking_ice_We need extensive cases for the *DE what they are bringing in and so fourth ..12:18
leitzTake your "Feature Set" or "Use Case" and test against that. With extra packagers able to parallel an expected outcome.12:19
wwoodsleitz: whoa. the point of bodhi voting is for testers to be able to say "pass/fail" 12:20
wwoodsif packages had test plans, this would be natural12:20
wwoodsbut we're kind of mixing up two different things: updates to stable releases12:20
wwoodsand new features going into the development tree12:20
viking_ice_There are potential privacy concerns regarding a feature of the user-switcher-applet it automatically logs you in on your IM server12:20
wwoodsnew features never hit bodhi during development12:21
-!- mether [n=sundaram@] has quit [Client Quit]12:21
leitzRight, but you're pas/failing at the package level.12:21
wwoodsviking_ice_: right, so the spec should say that can be disabled12:21
-!- mether [n=sundaram@] has joined #fedora-meeting12:21
wwoodsleitz: yeah - in theory each Feature has one or more packages, and each Feature has a test plan12:22
-!- drago01 [n=linux@chello062178124130.3.13.univie.teleweb.at] has quit [Remote closed the connection]12:22
wwoodsso once you get to a stable release, if you're updating one of the packages from a Feature12:22
* leitz is always subject to correction. :)12:22
wwoodsthen you can use that Feature's test plan to test the package12:22
viking_ice_wwoods: well I think the board needs to take a stand on that one should it be on by default or off I personally dont like some IM server know when I'm online or not and have some app set the status for me12:23
hadessdid i miss the fingerprint reader chat? poke me when you guys are ready to talk about it12:23
wwoodshadess: heh, sorry, QA meeting running long12:23
f13if we're into feature talk, I'm goign to step out and go have breakfast12:23
wwoodsFESCo should be at 1PM EST (1800UTC).. unless.. I'm mistaken?12:23
bpeppleno your right.12:24
leitzI'm suggesting not pass/failing by Feature because the understanding (subject to facts) is that there are ways to game the system.12:24
leitzSorry, /the understanding/my understanding/12:24
f13bodhi wouldn't be used for features.12:24
wwoodsleitz: I'm not sure I understand you, but currently Features and bodhi are completely separate12:24
wwoodsthe feature planning stuff only applies to Rawhide, which doesn't go through bodhi12:24
hadessWednesday at 17:00 UTC (13:00 EST) <- huh?12:25
leitzHmm...lemme go clarify my thinking and review my notes.12:25
wwoodshuh, it's listed as 1700-190012:25
-!- knurd_afk is now known as knurd12:25
hadess18:00 utc then12:25
wwoodscrap, are we stomping on FESCo?12:25
wwoodswe can take the rest of this discussion to #fedora-qa12:25
bpepplewwoods: no, we don't start for another 30 minutes or so.12:26
wwoodswell, either way, I need to get some lunch before I fall over12:26
wwoodsQA discussion will move to #fedora-qa12:26
f13and I some breakfast12:26
wwoodswe'll try to straighten out bodhi vs. features a bit12:27
* leitz will smooch with his wife for a while. :)12:27
viking_ice_soon supper for me....12:27
hadessbpepple: you got one or the other times wrong then12:27
wwoodsleitz, viking_ice_, f13, jlaska, jds2001: thanks. and anyone else I missed.12:28
bpeppleI probably forgot to adjust the UTC time. Big fail on my part obviously. :(12:28
-!- leitz is now known as leitz_afk12:28
-!- wwoods changed the topic of #fedora-meeting to: Channel is used by various Fedora groups and committees for their regular meetings | Note that meetings often get logged | For questions about using Fedora please ask in #fedora | See http://fedoraproject.org/wiki/Communicate/FedoraMeetingChannel for meeting schedule12:28
--- Log closed Wed Dec 10 12:28:12 2008

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!