Thursday, 2016-06-16

*** openstack has joined #openstack-swift05:59
*** dmorita has joined #openstack-swift06:05
*** dmorita has quit IRC06:10
*** ozeri has joined #openstack-swift06:12
*** SkyRocknRoll has joined #openstack-swift06:12
*** SkyRocknRoll has quit IRC06:18
*** klamath has quit IRC06:19
*** klamath has joined #openstack-swift06:19
*** rcernin has joined #openstack-swift06:24
kota_mahatic_: Did functests all passed?06:53
kota_mahatic_: i noticed one thing it could be a bug for container update EC + encyrption.06:54
kota_but I'm not sure right now.06:54
kota_hmm... current patchs in the chain are all fresh?06:55
kota_gerrit ui is hard to track the dependencies....06:55
kota_oh what...x-backend-container-override-etag is stale at encyrption-2!?06:58
kota_My concren wih -1 at encryption-1 might be meaningless :/06:58
kota_need deeper look anyways.06:59
kota_huh, it looks like x-object-sysmeta-container-update-override can overwrite the x-backend-container-override value.07:04
kota_it looks magic but it's right thing?07:05
*** tesseract has joined #openstack-swift07:09
mahatic_kota_: let me check that again. that issue (of tests passing on EC and not replication on a diff dev set up, i'll run them again.07:09
mahatic_kota_: I don't have tests failing when I download encryption-507:10
kota_mahatic_: thanks!07:10
mahatic_I downloaded that a couple of hours ago and ran functests07:10
kota_k07:10
kota_mahatic_: quick question07:10
kota_mahatic_: could you let me know if EC at the encryption-5 uses which header of sysmeta or historical one?07:11
kota_i mean x-object-sysmeta-container-update-override vs x-backend-container-override07:12
*** jamielennox is now known as jamielennox|away07:12
kota_maybe I'm dreaming or I don't know anything for encryption changes :/07:13
mahatic_kota_: :) Hold on, I'm checking that07:14
mahatic_kota_: it is the first one -  x-object-sysmeta-container-update-override, I don't see latter being used at all07:15
kota_mahatic_: thanks a lot!07:16
mahatic_this on encryption-5 patch07:16
kota_I'm at encryption-2 and it still uses later one.07:16
kota_so probably we have the change around 3-4.07:16
kota_gotcha, so my concen might be resolved around 3-4, not sure though.07:17
*** pcaruana has joined #openstack-swift07:17
kota_mahatic_: thanks for helping \o/07:18
openstackgerritOr Ozeri proposed openstack/swift: Raise 412 response on expirer  https://review.openstack.org/32690307:18
*** permalac has quit IRC07:19
kota_acoles: my last comment for patch 328204 might be addressed in later patch set. I'll try to look at them so if you have consideration for that already. it's worth that you let me know it.07:19
patchbotkota_: https://review.openstack.org/#/c/328204/ - swift (feature/crypto-review) - Support for http footers - Replication and EC07:19
mahatic_kota_: np. I think we will still have x-backend-container-override around. The commit msg in patch 328205 mentions that as well07:21
patchbotmahatic_: https://review.openstack.org/#/c/328205/ - swift (feature/crypto-review) - Make container update override headers persistent07:21
mahatic_kota_: if you download encryption-5 and takea  look at git log, the dependencies are easier to understand07:21
kota_ok, thanks!07:22
*** silor has joined #openstack-swift07:25
mahatic_kota_: sorry. I see both headers you mentioned in encryption-5 patch. (X-Backend-Container-Update-Override and ' x-object-sysmeta-container-update-override'. And it is correct that way if you look at dependencies07:26
kota_hmm07:27
*** rledisez has joined #openstack-swift07:28
mahatic_I have earlier searched with the header you mentioned which is missing 'update' word in the header07:28
mahatic_kota_: does it make sense? do you still have confusion on dependencies?07:28
kota_yup, thanks.07:29
*** geaaru has joined #openstack-swift07:29
kota_ah, I meant, no confusion for the dependencies :)07:30
mahatic_alright :) it's on the reverse 5 depends on 4, 4 on 3 etc07:32
*** silor1 has joined #openstack-swift07:33
*** silor has quit IRC07:34
*** silor1 is now known as silor07:34
*** klrmn has quit IRC07:40
*** hosanai has quit IRC07:51
*** SkyRocknRoll has joined #openstack-swift07:51
*** acoles_ is now known as acoles08:01
acoleskota_: do you mean this comment? https://review.openstack.org/#/c/328204/3/test/unit/proxy/controllers/test_obj.py@179108:07
patchbotacoles: patch 328204 - swift (feature/crypto-review) - Support for http footers - Replication and EC08:07
*** dmorita has joined #openstack-swift08:07
*** remix_auei is now known as remix_tj08:09
*** permalac has joined #openstack-swift08:09
openstackgerritDavanum Srinivas (dims) proposed openstack/swift: [WIP] Testing latest u-c  https://review.openstack.org/31844108:10
openstackgerritDavanum Srinivas (dims) proposed openstack/swift: [WIP] Testing latest u-c  https://review.openstack.org/31844108:10
*** dmorita has quit IRC08:11
permalacHello guys, is there any way to monitor Swift status with a GUI or WEB or ncurses ?08:12
acoleskota_: proxy will conitnue to use   x-backend-container-update-override headers, encrypter will use x-object-sysmeta-container-update-override headers. In the object server the encrypter headers will be preferred over the EC headers https://review.openstack.org/#/c/328205/4/swift/obj/server.py08:13
patchbotacoles: patch 328205 - swift (feature/crypto-review) - Make container update override headers persistent08:13
openstackgerritOr Ozeri proposed openstack/swift: pickle_async_update should create tmp_dir  https://review.openstack.org/33036308:15
admin6Hi there, I’m running some performance test around swift and I have two problems regarding latency to access "small" objects (for example 4KB)08:15
admin6First one is, when i GET a 4K object on my my current NAS over NFS storage, it takes 2ms, but the same object on swift takes 12ms. So instead of the usual 2 second to load a page with 1000 consecutive objetcs, it will theoretically takes 12 seconds…08:16
admin6Second problem, where the average latency is around 12ms, about 5% of my requests sucks and takes very high latency (around 500 to 1000ms). So my 1000 consecutive objects really takes around 30 sec to load…08:16
admin6Do you have any good suggestion or links I could follow to enhance these two problems?08:18
*** hseipp has joined #openstack-swift08:18
*** ma9 has joined #openstack-swift08:28
*** ma9 has left #openstack-swift08:32
*** jordanP has joined #openstack-swift08:34
kota_back from another meeting.08:37
kota_acoles: yes, I'm making sure how do i think of https://review.openstack.org/#/c/328204/3/test/unit/proxy/controllers/test_obj.py@179108:39
patchbotkota_: patch 328204 - swift (feature/crypto-review) - Support for http footers - Replication and EC08:39
kota_acoes: alright, you know, x-object-sysmeta-container-update is preferred x-backend-container-overrides08:40
kota_so it means it's ok any middlereware cannot overwrite *x-backend-container-override-value*.08:42
kota_use x-object-sesymeta-container-update-override instead.08:43
*** jamie_h has joined #openstack-swift08:43
acoleskota_: well, it means that encryption is not broken :) but I think your comment here is correct https://review.openstack.org/#/c/328204/3/swift/proxy/controllers/obj.py@214708:43
patchbotacoles: patch 328204 - swift (feature/crypto-review) - Support for http footers - Replication and EC08:43
kota_that makes me sense, encryption still works08:43
acoleskota_: any other middleware using x-backend-container-override-value would be broken by my change, as you say in the comment08:44
acoleskota_: do you know of any other middleware that uses x-backend-container-override- ?08:44
kota_acoles: nothing AFAIK ;)08:44
acoleskota_: so to be safe (for unknown middlewares) I should only overwrite x-object-sysmeta-ec- values that may come from the footers callback, but allow the footers to set the x-backend-container-override- values. good catch!08:46
kota_acoles: thanks, I tought I was just a negative worrier :/08:47
acoleskota_: that is the best kind of reviewer! ;)08:48
acolesthink I might need some more tests too08:48
kota_hew, great.08:50
*** dmk0202 has joined #openstack-swift09:02
*** ouchkernel has quit IRC09:02
*** ouchkernel has joined #openstack-swift09:05
*** jmccarthy has joined #openstack-swift09:24
kmARCacoles, notmyname: cca mathiasb and /me will review the crypto-related parts this afternoon around 5pm (UTC 3pm), will update you in the evening (and of course will leave comments on gerrit if needed)09:30
*** klrmn has joined #openstack-swift09:31
acoleskmARC: mathiasb: thanks! please leave any comments on gerrit.09:31
*** dmorita has joined #openstack-swift09:50
*** dmorita has quit IRC09:54
*** jmccarthy1 has joined #openstack-swift10:07
*** klrmn has quit IRC10:08
*** jmccarthy has quit IRC10:08
*** jamielennox|away is now known as jamielennox10:15
*** mvk_ has quit IRC10:26
*** manous has joined #openstack-swift10:33
*** mvk_ has joined #openstack-swift10:53
*** cdelatte has joined #openstack-swift10:57
*** kei_yama has quit IRC11:00
*** SkyRocknRoll has quit IRC11:39
*** McMurlock1 has joined #openstack-swift11:43
*** dmorita has joined #openstack-swift11:50
*** dmorita has quit IRC11:55
*** SkyRocknRoll has joined #openstack-swift11:56
*** rcernin has quit IRC12:24
*** pauloewerton has joined #openstack-swift12:35
*** rcernin has joined #openstack-swift12:38
*** mwheckmann has joined #openstack-swift12:40
*** zul_ has joined #openstack-swift12:41
*** zul_ has quit IRC12:42
*** zul_ has joined #openstack-swift12:42
*** zul_ has quit IRC12:43
*** zul_ has joined #openstack-swift12:44
*** zul_ has quit IRC12:49
*** zul has quit IRC12:49
*** zul has joined #openstack-swift12:50
*** rcernin has quit IRC12:54
*** klamath_ has joined #openstack-swift12:57
*** klamath has quit IRC12:57
*** hseipp has quit IRC13:02
*** SkyRocknRoll has quit IRC13:05
*** rcernin has joined #openstack-swift13:07
*** ppai has quit IRC13:17
*** ozeri has quit IRC13:26
*** raildo-a` is now known as raildo13:27
*** diogogmt has quit IRC13:29
*** diogogmt has joined #openstack-swift13:31
*** furlongm_ is now known as furlongm13:41
*** dmorita has joined #openstack-swift13:44
*** hseipp has joined #openstack-swift13:45
*** dmorita has quit IRC13:48
*** manous has quit IRC14:00
*** diogogmt has quit IRC14:01
*** sheel has quit IRC14:05
*** arch-nemesis has joined #openstack-swift14:08
*** arch-nemesis has quit IRC14:08
*** arch-nemesis has joined #openstack-swift14:09
*** daemontool has joined #openstack-swift14:19
*** jistr is now known as jistr|mtg14:28
*** sheel has joined #openstack-swift14:30
*** pcaruana has quit IRC14:41
*** diogogmt has joined #openstack-swift14:55
*** McMurlock1 has quit IRC14:56
*** psachin has quit IRC14:57
*** klrmn has joined #openstack-swift15:09
*** ouchkernel has quit IRC15:11
*** suyash has joined #openstack-swift15:15
*** ouchkernel has joined #openstack-swift15:17
*** _JZ_ has joined #openstack-swift15:19
*** jistr|mtg is now known as jistr15:19
*** silor has quit IRC15:21
*** dmorita has joined #openstack-swift15:26
*** dmorita has quit IRC15:26
*** dmorita has joined #openstack-swift15:26
*** klrmn has quit IRC15:29
*** zaitcev has joined #openstack-swift15:30
*** ChanServ sets mode: +v zaitcev15:30
*** dmorita has quit IRC15:31
openstackgerritAnne Gentle proposed openstack/swift: Adds migrated API reference files  https://review.openstack.org/31231515:32
*** vinsh has quit IRC15:32
*** vinsh has joined #openstack-swift15:33
*** openstackgerrit has quit IRC15:34
*** openstackgerrit has joined #openstack-swift15:34
*** vinsh_ has joined #openstack-swift15:40
*** vinsh_ has quit IRC15:40
*** vinsh_ has joined #openstack-swift15:41
*** nadeem has joined #openstack-swift15:41
*** nadeem has quit IRC15:42
*** nadeem has joined #openstack-swift15:42
*** vinsh has quit IRC15:44
acolesnotmyname: jrichli I have added a list to the trello board for questions that arise from crypto-review that may warrant broader discussion https://trello.com/b/63l5zQhq/swift-encryption15:46
jrichliacoles: thanks15:46
acolesit's hard to track those on gerrit, the comments get lost once new versions of a patch are pushed15:46
*** vinsh has joined #openstack-swift15:49
notmynameok, great15:53
notmynamelooking15:53
zaitcevCurrent Gerrit is better than old one, you can see the inline comments in the log.15:53
zaitcevI cannot stand Trello.15:53
*** vinsh_ has quit IRC15:53
notmynamebut with the review branch process where we end up pushing new patch sets, it's hard to find/see the comments on previous patch sets15:54
notmynameor at least easy to skip over them15:54
notmynamepesonally, I don't really care where it goes, as long as people have a way to find it15:56
*** garthb has joined #openstack-swift15:58
*** tesseract has quit IRC16:00
clarkbnotmyname: you can expand the list of comments and they should all be there. And if you toggle CI you can filter out all the CI comments16:03
clarkbshould give you a good thread of commentary per change16:03
notmynameclarkb: right. but since they're folded up by default and socially we're trained to look at the latest patch set's comments/info, it's easy to pass over comments on previous patch sets16:04
clarkbbut thats easy to correct right? "Reviewers please expand the comments and read the comment thread"16:06
*** gyee has joined #openstack-swift16:06
clarkbAlso I want to say gertty expands by default for those of you that are gertty users16:06
*** silor has joined #openstack-swift16:12
*** permalac has quit IRC16:15
*** silor1 has joined #openstack-swift16:18
openstackgerritAlistair Coles proposed openstack/swift: crypto - Test with non-zero offset  https://review.openstack.org/32702816:18
*** silor has quit IRC16:19
*** silor1 is now known as silor16:19
acolesmahatic_: jrichli I have merged content of patch 327028 into patch 328208 (next version)16:25
patchbotacoles: https://review.openstack.org/#/c/327028/ - swift (feature/crypto) - crypto - Test with non-zero offset16:26
patchbotacoles: https://review.openstack.org/#/c/328208/ - swift (feature/crypto-review) - Enable object body and metadata encryption16:26
jrichliacoles: nice16:26
*** lyrrad has joined #openstack-swift16:26
*** silor has quit IRC16:27
mahatic_acoles: thanks16:29
*** jordanP has quit IRC16:37
*** suyash has quit IRC16:41
*** dmk0202 has quit IRC16:42
timburkegood morning16:45
openstackgerritAlistair Coles proposed openstack/swift: Support for http footers - Replication and EC  https://review.openstack.org/32820416:56
openstackgerritAlistair Coles proposed openstack/swift: Make container update override headers persistent  https://review.openstack.org/32820516:57
openstackgerritAlistair Coles proposed openstack/swift: Enable middleware to set metadata on object POST  https://review.openstack.org/32820616:57
openstackgerritAlistair Coles proposed openstack/swift: Allow middleware to override metadata header checking  https://review.openstack.org/32820716:57
openstackgerritAlistair Coles proposed openstack/swift: Enable object body and metadata encryption  https://review.openstack.org/32820816:57
openstackgerritAlistair Coles proposed openstack/swift: Add encryption overview doc  https://review.openstack.org/32820916:57
*** suyash has joined #openstack-swift17:02
acoles^^ that set of patches should be up to date with all the review comments so far17:05
*** hseipp has quit IRC17:06
acolestimburke: good morning17:06
*** mvk_ has quit IRC17:06
*** daemontool has quit IRC17:08
notmynameacoles: thanks17:09
acolestimburke: about third party middlewares that may use container update override headers... I realised today that they would not override EC overrides on master, because EC sends them as footers (which middleware cannot do on master) and the object server prefers the footers over the headers.17:09
timburkemmm, good point17:10
acolestimburke: they'd work with replicated policies I think, but as mentioned before the POST container updates would be wonky.17:10
acolestimburke: do you know of any such middlewares?17:10
timburkeprobably not worth worrying about them, then. i do not know of any; i'm just paranoid about people actually *using* the features we write ;-)17:11
*** rledisez has quit IRC17:14
acolestimburke: yep, always good to think about that. I think we've done our best to accomodate them if they are out there (or at least warn of upgrade impact)17:14
acolestimburke: I made another change today in patch 328205 to how the footers callback metadata is applied w.r.t. EC to allow any such middleware to send old style container overrides in footers, but now realise that's moot since no middleware could be doing that with master. But doesn't hurt and prompted me to write some more explicit tests for that footer handling.17:17
patchbotacoles: https://review.openstack.org/#/c/328205/ - swift (feature/crypto-review) - Make container update override headers persistent17:17
acoles^^ that was based on kota spotting the issue17:17
*** rcernin has quit IRC17:21
*** klrmn has joined #openstack-swift17:23
*** tqtran has joined #openstack-swift17:31
*** dmorita has joined #openstack-swift17:33
*** ChubYann has joined #openstack-swift17:38
*** acoles is now known as acoles_17:39
*** mvk_ has joined #openstack-swift17:39
openstackgerritMichael Barton proposed openstack/swift: go: quarantine invalid files when replicating  https://review.openstack.org/33072717:47
*** rcernin has joined #openstack-swift18:07
*** geaaru has quit IRC18:09
*** gyee has quit IRC18:25
*** dmk0202 has joined #openstack-swift18:44
*** dmk0202 has quit IRC18:45
*** amit213 has joined #openstack-swift18:51
*** Jeffrey4l has quit IRC19:30
*** Jeffrey4l has joined #openstack-swift19:31
*** dmk0202 has joined #openstack-swift19:40
*** dmk0202 has quit IRC19:43
*** djc__ has joined #openstack-swift19:52
djc__why is the default keystone token expiration set to 24 hours? what are the ramifications of increasing beyond 24 hours?19:53
notmynamedjc__: default is 1 hour, these days, IIRC19:53
notmynamethe risk is that if a token is compromised, you only have to worry about it for the expiry period19:53
djc__notmyname: is there a reason for the 1 hour default? arbritary?19:54
djc__notmyname: ok..that's what I thought but wanted to confirm.19:54
notmynameyou'd have to ask the keystone team about it. it's set in keystone, not swift. you can find that team in #openstack-keystone19:54
djc__notmyname: posed the question there but didn't get a response. Thought someone here would know as well.19:55
*** admin6 has quit IRC19:56
notmynamedjc__: are you using swift and keystone today?19:58
djc__notmyname: yes. we have been running it for a while. We have it set to 24 hours but may need to reconsider that.20:01
notmynamecool20:01
djc__notmyname: does the 1 hour expiration pose a problem for transfers that last longer than 1 hour?20:03
notmynameno20:03
notmynamevalidation is done at the time of request20:03
notmynameyour apps should be reusing tokens instead of getting a new token for every request, so the token expiry is just about how often the app needs to get a new token20:04
djc__notmyname: thank you20:04
notmynameso if you set it to 1 hour instead of 24 hours, it means you'll be doing 24x the number of requests to the auth system every day, in an ideal setup20:05
notmyname24x sounds bad, but once an hour should be an issue :-)20:05
*** sheel has quit IRC20:05
*** djc__ has quit IRC20:13
*** openstackstatus has joined #openstack-swift20:20
*** ChanServ sets mode: +v openstackstatus20:20
*** zul has quit IRC20:21
*** zul has joined #openstack-swift20:22
*** dmk0202 has joined #openstack-swift20:23
*** gyee has joined #openstack-swift20:24
*** ouchkernel has quit IRC20:29
*** mwheckmann has quit IRC20:32
*** ouchkernel has joined #openstack-swift20:37
*** arch-nemesis has quit IRC20:41
*** zul has quit IRC20:52
*** arch-nemesis has joined #openstack-swift20:56
*** jamie_h has quit IRC20:57
*** arch-nemesis has quit IRC21:03
openstackgerritMerged openstack/swift: crypto - Test with non-zero offset  https://review.openstack.org/32702821:15
*** pauloewerton has quit IRC21:15
*** arch-nemesis has joined #openstack-swift21:17
*** diogogmt has quit IRC21:34
*** diogogmt has joined #openstack-swift21:36
*** rcernin has quit IRC21:36
*** dmk0202 has quit IRC21:38
*** dmk0202 has joined #openstack-swift22:04
*** lcurtis has joined #openstack-swift22:08
*** darrenc is now known as darren_afk22:34
*** dmk0202 has quit IRC22:39
*** lcurtis has quit IRC22:54
*** darren_afk is now known as darrenc22:55
*** _JZ_ has quit IRC23:07
*** diogogmt has quit IRC23:10
*** garthb has quit IRC23:20
*** nadeem has quit IRC23:24
*** chlong has quit IRC23:42

Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!