Wednesday, 2013-09-25

*** acabrera has joined #openstack-marconi00:14
*** acabrera has left #openstack-marconi00:15
*** nosnos has joined #openstack-marconi00:16
*** kgriffs is now known as kgriffs_afk00:20
*** reed has quit IRC00:24
*** acabrera has joined #openstack-marconi00:26
*** ayoung has quit IRC00:27
*** acabrera has left #openstack-marconi00:27
*** oz_akan_ has joined #openstack-marconi00:35
*** malini is now known as malini_afk00:43
*** tedross has joined #openstack-marconi01:26
*** oz_akan_ has quit IRC01:49
*** oz_akan_ has joined #openstack-marconi01:49
*** oz_akan_ has quit IRC01:51
*** oz_akan_ has joined #openstack-marconi01:51
*** tedross has quit IRC01:58
*** oz_akan_ has quit IRC02:25
*** oz_akan_ has joined #openstack-marconi02:26
*** oz_akan_ has quit IRC02:30
*** oz_akan_ has joined #openstack-marconi03:36
*** nosnos has quit IRC03:40
*** nosnos has joined #openstack-marconi03:41
*** oz_akan_ has quit IRC03:42
*** nosnos has quit IRC03:42
*** nosnos has joined #openstack-marconi03:44
*** kgriffs_afk is now known as kgriffs03:48
*** reed has joined #openstack-marconi03:58
*** kgriffs is now known as kgriffs_afk04:01
*** reed has quit IRC05:07
*** flaper87|afk is now known as flaper8707:30
*** ykaplan has joined #openstack-marconi07:51
*** gordonsim has joined #openstack-marconi08:00
openstackgerritA change was merged to openstack/marconi: fix(mongodb): Remove $or clause used to claim messages
*** ykaplan has quit IRC10:59
*** oz_akan_ has joined #openstack-marconi11:20
*** ykaplan has joined #openstack-marconi11:27
*** oz_akan_ has quit IRC11:38
*** ykaplan has quit IRC11:40
*** ykaplan has joined #openstack-marconi11:44
flaper87here it comes!11:59
openstackgerritFlavio Percoco proposed a change to openstack/python-marconiclient: Restructure client code
flaper87there it is11:59
flaper87Clien tests take 0.00s each! w000t12:00
*** tedross has joined #openstack-marconi12:12
openstackgerritFlavio Percoco proposed a change to openstack/python-marconiclient: Add common http class
*** acabrera has joined #openstack-marconi12:40
*** acabrera is now known as alcabrera12:41
alcabreraMorning! :)12:41
openstackgerritFlavio Percoco proposed a change to openstack/python-marconiclient: Add common http class
flaper87alcabrera: yo yo12:45
openstackgerritFlavio Percoco proposed a change to openstack/python-marconiclient: Restructure client code
*** oz_akan_ has joined #openstack-marconi12:49
*** oz_akan_ has quit IRC12:52
*** oz_akan_ has joined #openstack-marconi12:52
alcabreraflaper87: Those 2 patches have been reviewed - +2 on both of them.12:53
alcabreraI'm going to look at the rest of the review queue now. :D12:53
flaper87alcabrera: AWESOME! Thanks!12:54
flaper87btw, I replied to one of your comments in one of my patches :D12:54
alcabreraMoving the unit tests - yeah, I saw that one! I agree with you entirely. I reviewed so much yesterday I lost track of what was doing what. :P12:55
flaper87No worries at all. :)12:56
*** ykaplan has quit IRC12:57
alcabreraflaper87: I'll be filing that proxy bug once I go through reviews.12:57
alcabreraA bug to address all the standing design issues - there's a few left yet. :)12:58
alcabreraThough I did file the bug for eliminating the global config.12:58
flaper87alcabrera: yeah, saw that bug, thanks! I actually picked it :P13:01
flaper87alcabrera: if they are not so many, a bug for each pending issue would be awesome. That would help to get folks to contribute on those13:01
alcabreraflaper87: +113:02
alcabreraflaper87: ah, what do you think of adding :type: and :rtype: to docstrings?13:02
alcabreraI noticed sphinx has this feature yesterday.13:02
flaper87alcabrera: +1 lets enforce that in reviews13:03
alcabrerasweet. I started doing it rather adhoc-style, e.g., `:param hosts: [text] - a list of URLs`, but that was before I realized the type could be specified in a special way.13:04
flaper87btw, a gentle reminder, I think there are some patches that were submitted before the structure refactor, pls, double check that those patches don't add things following the old structure13:04
flaper87I did it but you know, just in case13:04
flaper87I mean, if you happen to review one of those patches13:05
alcabreraflaper87: I'll keep an eye out for patches that might violate the restructuring.13:05
flaper87alcabrera: danke!13:07
*** ayoung has joined #openstack-marconi13:10
*** mpanetta has joined #openstack-marconi13:12
openstackgerritAlejandro Cabrera proposed a change to openstack/marconi: proxy: adds oslo.cache, config support, & stevedore
flaper87who's this d00d? ^13:18
openstackgerritAlejandro Cabrera proposed a change to openstack/marconi: fix(proxy): forward marconi headers
openstackgerritAlejandro Cabrera proposed a change to openstack/marconi: proxy: mirror structure of marconi queues + bootstrap
openstackgerritAlejandro Cabrera proposed a change to openstack/marconi: test(proxy): fix proxy-transport tests
alcabreraall the rebasing~13:19
*** oz_akan_ has quit IRC13:24
*** oz_akan_ has joined #openstack-marconi13:24
*** kgriffs_afk is now known as kgriffs13:25
kgriffsat the airport13:25
mpanettakgriffs: On your way here?13:25
kgriffsonline thanks to the nation's largest wifi network: NETGEAR13:26
mpanettaWhich airport?13:26
flaper87kgriffs: have a safe flight13:26
mpanettaAh cool13:26
mpanettaWatch the weather on your way.13:26
kgriffsi've got about 45 min before I have to board13:26
mpanettaI'm about to attempt to cherry pick this patch.13:27
openstackgerritAlejandro Cabrera proposed a change to openstack/marconi: proxy: mirror structure of marconi queues + bootstrap
flaper87kgriffs: then is waiting for you :D13:29
alcabrerakgriffs: o/13:29
mpanettaOr maybe not.13:33
*** jcru has joined #openstack-marconi13:37
kgriffsflaper87: :D13:37
alcabrerakgriffs: here's some amusing pre-flight reading:
alcabreraI happened upon that this morning and found some very amusing bits of technical wordplay. :P13:40
* alcabrera has now reviewed all marconi-related patches.13:41
alcabreraNow for some bug reports13:41
alcabrerapy3k is now a voting gate for marconiclient.13:42
kgriffs"MongoDB adds a "PLEASE" keyword for inserts, boosting chance that data is stored to above 75%."13:43
kgriffsit's FUD, but still funny13:43
alcabrerakgriffs: That's what I'm talking about, haha.13:43
kgriffs(mongodb has had good persistence for quite a while now)13:43
* kgriffs follows13:44
alcabrera'A load balancer walks into a shareholder meeting. CEO says "Nope, no proxies allowed."'13:44
alcabrerahmmmm ^13:45
* kgriffs things the real TechCrunch is evil13:45
alcabrera"BREAKING – Dropbox and Box agree to merger, new company to be called DropboxBox"13:45
kgriffswhile we are on a tangent, you guys folllow @wwwtxt ?13:45
flaper87kgriffs: pleeeasseeeee13:45
kgriffsit's pretty darn awesome13:46
alcabrerakgriffs: not yet - I'll check it out. :D13:46
flaper87mmh, dunno any wwwtxt13:46
kgriffsflaper87: not now, I'm busy wasting time on the internetz!13:46
kgriffsIt's some guys thesis - he tweets old text from the intertubes that is eerily relevant to current events.13:47
flaper87kgriffs: LOOOL13:47
alcabreraThe Apple-related ones are quite - well, it reflects the state of things pretty well. :P13:47
kgriffsflaper87: can you just add a NOTE there:
kgriffsthen I will +213:48
kgriffsalso, moving it outside the try block may help reduce confusion13:48
kgriffs(I am assuming you don't expect L30 to fail)13:48
flaper87kgriffs: you're right, lemme do that13:49
flaper87don't jump on that plain yet13:49
kgriffsre breaking the dep between those patches, TBH, I'm not convinced it is worth the trouble making them independent right now. Some of the lines overlap and whichever merges first will cause some extra rebasing work with the other13:49
kgriffsre benchmarking them, if we benchmark just the marker gen/counter patch13:50
kgriffsthen benchmark the same with p+q added on top, we should be able to extrapolate over the relative performance gain/loss between the two13:50
kgriffsI think the the marker patch needs to go in regardless so we can get rid of marconi-gc13:51
kgriffsso, it is about to become the new baseline anyway13:51
flaper87sounds like a plan to me! My main concern is the benchmark results we saw, I wanted to let the q+p patch land first instead of keeping it around waiting for the marker patch to land13:52
flaper87I like the marker patch, TBH13:52
kgriffsbtw - i was considering doing the p+q treatment on the queues collection as well, since that is now used to lookup/increment the counter when posting messages13:52
kgriffsflaper87: I am about to address your last comment, and if there is nothing else, the marker patch is good to go, nicht?13:52
alcabrerakgriffs: I'm +1 on the marker patch. Even if there's a slight drop in perf, I think the new approach is easier to reason about.13:52
flaper87+1 for me13:53
kgriffsthere isn't a drop in perf13:53
flaper87from me13:53
alcabrerakgriffs: yeah, unify the storage layer across the p+q treatment. It's likely relevant to all the other storage drivers (future and current), as well.13:54
kgriffsand the test doesn't account for GC time which should be significantly improved, as well as retries on marker collision13:54
openstackgerritFlavio Percoco proposed a change to openstack/python-marconiclient: Authentication API + NoAuth backend
openstackgerritFlavio Percoco proposed a change to openstack/python-marconiclient: Add keystone backend
alcabreraIt also helps a lot that we reached consensus on the whole delimiter thing - "project/queue"13:54
kgriffsflaper87: thoughts on p+q in queues collection?13:54
flaper87kgriffs: alcabrera guys, could you re+2 those 2 patches?13:54
flaper87kgriffs: +113:55
flaper87that will also keep some consistency between both resources13:55
kgriffsyes, just as soon as gerrit +1s it. ;)13:55
flaper87I already gave it some nutella, he'll +1 it13:56
flaper87ok, gerrit +1'd the first13:56
alcabreraflaper87: +2'd13:56
flaper87alcabrera: danke13:56
alcabreraI double-checked the py3 details in the process, to make sure those stayed healthy. :P13:56
flaper87and the second one!13:57
* alcabrera cringes whenever he sees print statements13:57
flaper87alcabrera: Py3K is now voting :D13:57
alcabreraflaper87: I saw the email! :)13:57
flaper87meaning, we can now trust jenkins +1 on Py3k support13:57
alcabreraflaper87: as long as our test coverage stays decent, of course13:58
alcabrera(which it will)13:58
openstackgerritA change was merged to openstack/python-marconiclient: Authentication API + NoAuth backend
flaper87alcabrera: LOOOOOOOOOOOOL13:58
flaper87kgriffs: the second one got +113:58
kgriffsjust thinking ahead, eventually there won't be a 1:1 mapping between project ID and auth token, as in a single token will be authorized for multiple project IDs. We will need a way then to set X-Project-ID in the client. That will also require updating keystone middleware to stop clobbering X-Project-ID and instead verify it with belongsTo, but something to keep in mind for later.14:00
flaper87knock knock, still there?14:00
flaper87HOLY MOLY!14:00
flaper87that's a quesiton14:00
* flaper87 sits and reads14:01
* alcabrera notes that this is not a question14:01
alcabrerakgriffs: good point, though.14:01
flaper87alcabrera: LOL, just noted that!14:01
*** yassine has joined #openstack-marconi14:02
flaper87kgriffs: yeah, so, that code already adds a way to set the token either through CLI params or config options14:03
kgriffsok, so we can just follow that pattern to allow setting other auth headers14:03
flaper87if the token is manually set, it won't try to get one *unless* the request returns a 40114:03
flaper87and just then it'll retry to get a new one (if the user wants so)14:03
kgriffs(e.g., X-Project-ID)14:03
flaper87kgriffs: yeah!14:03
kgriffscool beans14:03
flaper87X-Project-ID is something the Transport instance will take care of14:04
flaper87but yes, you're right14:04
flaper87just trying to share what I have in mind14:04
kgriffsgotta board soon14:05
kgriffsquick question14:05
kgriffshow does that end up setting test-token?14:05
flaper87kgriffs: yup14:05
flaper87the header, that's what I meant to test there14:05
kgriffsthe mock in the other test sets it14:05
kgriffsbut no mock in test_with_token?14:06
flaper87kgriffs: yup because the first test mocks a keystone call14:06
flaper87the second one gets the token from configs14:06
kgriffsoh, duh14:06
kgriffsI'm rushing since I gotta go14:06
flaper87thank you!14:06
alcabrerakgriffs: enjoy the flight. :)14:06
flaper87no worries14:06
flaper87have a safe flight, after you +2 that patch14:07
alcabreraThat is the duty of a core reviewer - +2, -2; all other things are secondary.14:07
*** nosnos has quit IRC14:08
kgriffsall ur patch are belong to me14:10
openstackgerritA change was merged to openstack/python-marconiclient: Add keystone backend
flaper87kgriffs: thanks d000000d :D14:11
flaper87alcabrera: btw, didn't know hoist :D14:12
*** malini_afk is now known as malini14:13
kgriffsboarding any minute now14:14
alcabreraflaper87: I'd seen it around in the marconi code, mostly's. I had to verify that it was a word (google: define hoist) just before I responded. :)14:15
malinialcabrera: I am all fuming now with my cake story this morning & I see ur comment 'my tasty {0} cake'.format('chocolate')'14:19
maliniany example today should not involve cake :D14:19
alcabreramalini: hahaha14:20
alcabreramalini: After reading your email, I had cake on my mind14:20
alcabreravery relevant14:20
flaper87cake? Chocolate cake? cake?14:21
flaper87can I have some?14:21
maliniflaper87: its a chocolate cake & I waited for 40 loooong minutes waiting for the guy to finish it up!14:22
maliniwish I cud send u some..14:22
maliniu have a 3D printer? ;)14:22
flaper87malini: not sure if the taste would be the same, I'm pretty sure the consistency will, though14:22
malinialcabrera: w.r.t ur review comments, did you mean to have all the asserts in an internal function?14:25
alcabreraflaper87: on more than one occasion while going grocery shopping, I have though, "Man, if only I had Flavio's address! I'd totally send him a box of poptarts!". :D14:25
malinido they have poptarts in italy?14:26
alcabreramalini: for the tests test_get_stats_queue_no_claimed and test_get_stats_queue_one_claimed, yes.14:26
*** ametts has quit IRC14:26
*** kgriffs is now known as kgriffs_afk14:26
malinialcabrera: I am a lil wary abt that, since that will make it less obvious in the test about wht we are verifying14:27
maliniespecially since it is just a bunch of asserts14:28
alcabreramalini: If you could move the common parts of those two tests into probably three functions, that'd be ideal. Something like: _test_stats_match_expected(self, xclaimedcount=1, xfreecount=..., xtotal=...)14:28
flaper87alcabrera: I can give you my address :D14:28
flaper87wait, isn't it in google already?14:28
flaper87I'm surprissed google bike hasn't come yet14:29
alcabreramalini: check out (what do you think of this approach?)14:30
malinialcabrera: I would rather make it a new assert in the base class..Again I am not convinced abt moving it away from the tests, since we wont use it any other tests.14:30
malini looks like a great candidate for a new assert in the base class ;)14:31
alcabreraflaper87: yeah, send it to me in /msg or an email. I'll send something your way sometime. :)14:31
alcabreramalini: hehe, it may be the case. I thought it was pretty specific to the storage:partitions unit since it targets just the structure of that, but I think I see what you mean!14:32
malinialcabrera: like this
malini(it feels weird typing aLcabrera)14:32
alcabreramalini: Something like an a general @abstractmethod: test_check_structure or something like that.14:32
alcabreramalini: changed my name last night - acabrera was taken and I was tired of being cppcabrera, haha14:33
alcabreramalini: ahh, I see. The only problem I have with something like assertIsSubset in the general case is that there's the danger of matching up values incorrectly. I can see it working for headers alright, but for more mixed fields, I like having the option of matching them up by argument name, like here:
malinialcabrera: I can make those named arguments.14:43
alcabreramalini: for the assertIsSubset?14:47
* alcabrera got confused :(14:47
malinialcabrera: yes..14:47
alcabreramalini: I don't think assertIsSubset needs changes - it's *really* good at doing set comparisons.14:51
alcabreraBut bringing it back to the original point - the stats test patch (which is awesome!), I'm okay with merging the patch as is.14:52
alcabreraerr, +2-ing, anyway. :P14:53
malinidont +2 yet14:54
maliniI am fixing some stuff :D14:54
alcabreramalini: alright, let me know. :)14:55
alcabreramalini: when you're done, could you place a note in the review comments regarding what you fixed?14:55
malinialcabrera: sure14:56
*** cthulhup has joined #openstack-marconi14:57
*** amitgandhi has joined #openstack-marconi14:58
*** cthulhup has quit IRC15:00
*** cthulhup has joined #openstack-marconi15:01
*** fvollero has joined #openstack-marconi15:02
flaper87fvollero: yo, welcome!15:03
fvolleroflaper87: yo thanks bro :)15:03
flaper87guys, I'd like to introduce fvollero to you! He's a guy I found sitting outside my door begging me to do something reall really really cool!!! I invited him to join15:04
alcabrerafvollero: heeey! Welcome!15:04
fvolleroflaper87: Lol, funny as hell... :)15:04
* alcabrera notes that cool things do happen here15:04
fvolleroalcabrera: Thanks a bunch man!15:04
malinihello fvollero!!!15:04
fvolleromalini: hello :)15:04
flaper87jokes apart, fvollero is a fellow red hatter and he'll contribute as much as his time permits to Marconi15:05
* flaper87 gets LP bugs url15:05
flaper87muahahhahaha muahahha15:05
fvolleromeh :)15:06
maliniblasphemy!! flaper87 thinks doing cool stuff in marconi is a joke!!!!!15:06
maliniGlad to have you here fvollero !15:06
fvolleroflaper87: i'll try, as much is in my power... to not choke you or poison the sweets you desire so much15:06
fvolleromalini: let's wait to see some of my code before saying that :)15:06
flaper87malini: I never said that :( I said in my open-mic that Marconi is the coolest channel / team / project EVER!15:07
flaper87fvollero: there you go!15:08
flaper87that should give you some context about Marconi :)15:09
flaper87please, don't hesitate to ask anything!15:09
maliniflaper87: just kidding :)15:09
flaper87malini: you gotta give me a piece of that cake now!15:10
flaper87and don't play the 3D printer card!!!15:10
maliniI am sure ametts will let me expense the ticket to italy15:11
openstackgerritAlejandro Cabrera proposed a change to openstack/marconi: fix: age presented as negative
alcabreraspeaking of bugs ^^15:11
flaper87I moved a file in one of my recently approved commits that I now want to move back.... DON'T HATE MY SCHIZOPHRENIA15:13
alcabreraflaper87: lol15:15
alcabreraflaper87, malini: Could I get a review on the pending 'negative age' patch?15:15
flaper87alcabrera: Yes sir, right away15:15
malinialcabrera: am already on it15:16
flaper87alcabrera: reviewed15:18
flaper87you left a print in the test15:18
flaper87I'm ok with removing that other code in this patch, I don't think this patch will be reverted, ever.15:18
alcabreraheh, those print()s.15:18
alcabreraflaper87: cool15:18
openstackgerritAlejandro Cabrera proposed a change to openstack/marconi: fix: age presented as negative
alcabreraprint removed ^^15:19
flaper87alcabrera: awesome, LGTM. Lets wait for Jenkins to bless it15:19
malinialcabrera: I added a couple of comments in the first patch set15:20
alcabreramalini: thanks!15:20
alcabreramalini: yeah, I agree with your 'incr/wait' comment. I'd add that as a separate patch, though.15:21
alcabreraThankfully, the mongodb driver does age calculations based off of utcnow, so that'd be an easy test to write.15:22
maliniwe might also want to do the same test in claims15:22
alcabreraunfortunately, the sqlite driver doesn't use utcnow... :/15:22
flaper87I'd prefer not sleeping the test, TBH15:22
alcabreraflaper87: I'm with you.15:23
maliniflaper87: why ?15:23
flaper87malini: because it makes tests slow15:23
flaper87and we're just at the very beggining of Marconi's development15:23
malinido you have an alternate approach in such scenarios?15:24
flaper87there'll be tons of new tests and I'd prefer to mock things instead of sleeping15:24
fvolleroalcabrera: yeah, meanwhile in the real world, people write code :)15:24
flaper87malini: I guess patching utcnow works for mongodb's driver15:24
alcabreraflaper87: patching utcnow definitely works. I enjoyed seeing the -10 age when I first set out to fix this bug. :)15:25
alcabreraflaper87: for mongodb, that is.15:26
maliniflaper87: works for me15:26
flaper87TBH, in this very specific case, I'm not sure that test is actually worth it. I mean, it's all a `now - utils.blah_blah` calc after all.15:26
flaper87but hey, we managed to invert those 2 so...15:26
maliniI am very skeptic of 'this will never break' :d15:27
maliniThe last time I said 'it'll never happen in prod' - that was the first bug we got from support :D15:27
alcabreraheh. :P15:28
alcabreramalini: thankfully, this test does exercise marconi's production grade driver correctly.15:28
alcabrera**storage driver15:28
alcabrerathe sleep isn't necessary to trigger the age calculation15:28
malinimalini:  I am good with what we have, except adding the test for claim age too15:32
*** whenry has joined #openstack-marconi15:32
flaper87malini: you talking to yourself?15:33
maliniI missed my medication today ;)15:33
flaper87flaper87: I'm good with what we have.... flaper87 nooo we're not, you shouldn't be doing that... flaper87 yes, I'll and I'm doing so15:34
flaper87we should write a book out of what we talk in this channel15:34 can read minds , just like I do!!15:34
flaper87seriously, it's like a mix of techy insanity with real world examples15:35
alcabreramalini: oooohhh, I'm so glad I decided to test the claim deal15:37
alcabreramalini: claims *also* have the bug15:37
alcabreranow fixing15:37
alcabrera(negative age on claims)15:37
maliniheheee..I can think up bugs..I have the POWER!!15:37
alcabreramalini: :D :D :D15:37
alcabrera"MismatchError: -1 is not < -1015:37
flaper87malini: you know what? I'll give you my piece of cake, you deserve it!15:39
alcabreranow to run pep8 and...15:39
openstackgerritAlejandro Cabrera proposed a change to openstack/marconi: fix: age presented as negative
alcabreramalini, flaper87: ^^15:41
alcabreraThere ya go - squeaky clean marconi-queues.15:41
alcabreranot as tasty as cake - but it'll do. :P15:41
*** reed has joined #openstack-marconi15:42
maliniI think I am finally getting over the cake paranoia..I am no longer repulsed by cake15:43
alcabreraback - desk lunch, woot!15:51
*** ykaplan has joined #openstack-marconi15:53
openstackgerritFlavio Percoco proposed a change to openstack/python-marconiclient: Add a base transport class
openstackgerritFlavio Percoco proposed a change to openstack/python-marconiclient: Restructure client code
openstackgerritFlavio Percoco proposed a change to openstack/python-marconiclient: Add a base transport class
alcabreraflaper87: what changed in 48223 (restrcuture)?15:54
flaper87alcabrera: I moved transport/request back15:55
flaper87alcabrera: I moved transport/ back15:55
alcabreraoh, *that*. :P15:55
flaper87technically, I didn't touched it at all :D (at least git doesn't show that anymore)15:55
alcabreraI can see that now.15:55
* alcabrera counts the number of changed files15:56
flaper87ah fuck, I completely forgot to run tests for the last one15:56
flaper87SOOO ASHAMED15:56
alcabrerajenkins will know. >:)15:56
alcabreraTime to write those proxy bug reports~15:57
alcabreraThen, I need to tackle that admin API deal (before I address technical debt). D:15:57
fvolleronaive question, where i get the code ? :D16:01
alcabrerathe first is server-side, the other is client-side. :)16:01
*** yassine has quit IRC16:07
openstackgerritFlavio Percoco proposed a change to openstack/python-marconiclient: Add a base transport class
openstackgerritMalini Kamalambal proposed a change to openstack/marconi: Add Tests for Queue Stats
*** whenry has quit IRC16:21
alcabreraflaper87: marconi proxy bugs -
alcabreraI catalogued the most pending ones.16:22
flaper87alcabrera: man, you rock!16:22
flaper87thanks for that!16:22
* alcabrera mixes idioms all the time16:22
*** amitgandhi has quit IRC16:38
*** ykaplan has quit IRC17:16
alcabreraflaper87: I've figured out how to create separate config files now - I'm getting to understand oslo.config a little more each day!17:19
alcabreraI'm looking at having: ~/.marconi/[marconi, proxy].conf17:20
alcabreraperhaps that marconi.conf should be queues.conf...?17:20
flaper87alcabrera: I'd prefer marconi-(queues|proxy|notifications).conf17:21
alcabreraeven though they would be created in the .marconi directory?17:21
alcabreraadding the extra marconi- seems redundant, or am I missing something? :x17:22
alcabreraflaper87: ^17:22
flaper87well, our default folder is /etc/marconi or .marconi but, users can put those files wherever they want17:22
alcabreraI see.17:22
alcabrera~/ and /etc/ are on the default search path, after all17:23
alcabreraalright, +1 to marconi-X for X in {queues, notifications, proxy}17:23
alcabreraflaper87: I'm also agreeing with you more over time that the common.config needs to be eliminated - I had to add the prog parameter to it to get this scheme working. :P17:25
flaper87alcabrera: yeah, it also hides some nice methods in ConfigOpts (like find_config_file)17:26
flaper87so, either we completely proxy ConfigOpts or we eliminate common.config17:26
flaper87and I prefer the later17:26
alcabreraI prefer the latter, as well (atm). It seems like too much trouble to hide the interface behind something slightly friendly. Seems it'd be more effective to make the changes in oslo.config (2.0+) if the interface could benefit from being nicer.17:29
alcabrera(**slightly more friendly17:30
alcabreraflaper87: while on the subject of configs - for oslo.cache, I feel like having the config section named [oslo_cache] *and* having the cache_* prefix to the options is too redundant. Thoughts on simplifying to backend, prefix vs. cache_backend, cache_prefix?17:31
*** cthulhup has quit IRC17:34
flaper87alcabrera: ah yeah!!17:35
flaper87that sounds kinda redundant to me too17:35
flaper87mmh, also, the cache_prefix should actually be namespace17:36
flaper87and cache_backend could be default_backend17:36
alcabreraWant me to submit this as a bug report? :D17:36
alcabreraflaper87: I'm not sure about default_backend... seems like the choice of backend would be immutable after the application using oslo.cache has launched.17:37
flaper87alcabrera: add a comment here :
flaper87alcabrera: it is not immutable, in theory you could call get_cache as many times as you want17:38
alcabrerathat's true17:39
alcabreradefault_backend works for me, in that case.17:40
*** amitgandhi has joined #openstack-marconi17:47
malinialcabrera: will proxy authenticate with a regular x-auth-token, or is there a super user thing?18:09
alcabreramalini: regular x-auth-token18:10
alcabreramalini: the admin portion of the API is meant to be kept in a place that is not accessible by public users18:11
maliniq2. why don't we have an update partition?18:11
alcabreramalini: heh - I didn't think to add it. >.>18:12
alcabreramalini: which is actually a pretty big oversite18:12
maliniyeah..we might want to add a new host to the partition , remove one etc18:12
alcabreraCould you create a bug for that?18:12
alcabreraI've been prefixing all proxy-related bug titles with [proxy]18:13
alcabrerathat way, they're easy to find18:13
alcabreraGET,PUT /v1/partitions/{partition}/hosts and GET,PUT /v1/partitions/{partition}/weight should be a thing18:16
alcabrerathose are missing admin routes, IMO. ^18:16
alcabreramalini: gracias!18:26
malinialcabrera: de nada18:27
malinialcabrera: Since weight corresponds to probability, the sum for all partitions should always be 100, rt?18:32
alcabreramalini: no - weight is actually summed up to form a spectrum.18:35
alcabreraSo if we have a total weight of 500 between all partitions, and partition B has a weight of 50, it's chance of being chosen is 50/50018:35
maliniaah ..ok18:35
maliniit's a lil conter-intuitive. like with the response here , its hard to understand how often a partition will be used18:37
alcabreramalini: it takes a little more to get to the answer than if probabilities were used, I agree there. In that example, "wat3" would be chosen (25 / (256 + 500 + 25)) times.18:41
alcabreraerr, not times, with a probability of that. :P18:41
*** gordonsim has quit IRC18:41
alcabrerathe benefit to usig a weighted system is that we can have more than 100 partitions. With probabilities, we're capped at 100 (if we stuck with integral probabilities) since 1 * 100 = 100%18:42
alcabreraThe best reference to seeing the weighted algorithm in depth is:
malinitaht is right..or else with adding anew partition, you'll have to update all the old18:48
maliniI take my 100 probability back :)18:49
*** reed has quit IRC19:22
*** amitgandhi has quit IRC19:43
openstackgerritAlejandro Cabrera proposed a change to openstack/marconi: feat: separate config for queues and proxy
openstackgerritAlejandro Cabrera proposed a change to openstack/marconi: fix(proxy): forward marconi headers
*** malini is now known as malini_afk20:00
*** amitgandhi has joined #openstack-marconi20:04
openstackgerritAlejandro Cabrera proposed a change to openstack/marconi: feat: separate config for queues and proxy
alcabreraflaper87: Here's that patch that separates out configs in a by-project fashion.20:11
flaper87alcabrera: awesome, I'll take a look at it!20:12
flaper87Thanks for the hard work on the proxy code20:12
alcabreraflaper87: np - it continues to be fun. :)20:19
alcabreraflaper87: it may involve a touch of rebasing, but I plan to come back to the mirroring patch and to handle as much technical debt as possible there.20:20
alcabrera(given I need to finish the public/admin proxy API split first)20:20
*** flaper87 is now known as flaper87|afk21:03
*** alcabrera has quit IRC21:04
*** flaper87|afk is now known as flaper8721:13
*** oz_akan_ has quit IRC21:22
*** malini_afk is now known as malini21:26
*** malini is now known as malini_afk21:39
*** malini_afk is now known as malini21:47
*** amitgandhi has quit IRC21:54
*** mpanetta has quit IRC22:14
*** jcru has quit IRC22:20
*** tedross has quit IRC22:31
*** oz_akan_ has joined #openstack-marconi22:33
*** oz_akan_ has quit IRC22:38
*** reed has joined #openstack-marconi23:16
*** flaper87 is now known as flaper87|afk23:33
*** malini is now known as malini_afk23:37
*** malini_afk is now known as malini23:56

Generated by 2.14.0 by Marius Gedminas - find it at!