Thursday, 2015-05-07

*** DCWillia_ has quit IRC00:02
*** annegentle has joined #openstack-swift00:03
*** minwoob_ has quit IRC00:06
*** annegentle has quit IRC00:08
openstackgerritDenis Cavalcante proposed openstack/swift-specs: Internal communication measurement for swift spec  https://review.openstack.org/18080100:19
pelusetorgomatic, tricky00:23
notmynamemy favorite thing about EC: that it's "beta"00:27
*** dmorita has joined #openstack-swift00:28
mattoliveraunotmyname: lets make everything beta!!00:30
notmynamelol00:30
* mattoliverau has landed and found a coffee btw00:30
pelusethe only thing better than a beta is "a prototype" :)00:32
notmynamesummary for today is tracing down some bugs in both read and write. clayg / torgomatic have a patch for some PUT stuff00:35
notmynamestill tracking the GET stuff00:35
notmyname6+4 scheme, proxy finds 6 but still returns 503 to the client00:36
notmynameand other times there's a traceback from liberasure00:36
notmynameI don't know if they are the same or related00:36
*** annegentle has joined #openstack-swift00:39
*** DCWilliams_VA has joined #openstack-swift00:40
pelusehow to repro the GET issue?00:41
*** annegentle has quit IRC00:56
*** kota_ has joined #openstack-swift00:57
kota_good morning00:58
*** shakamunyi has quit IRC01:02
*** thurloat is now known as thurloat_isgone01:04
mattoliveraukota_: Morning!! How's the holidays?01:09
claygpeluse: dunno how to repro - but it is reproduable01:09
claygpeluse: notmyname had this weird object with the fragments all mixed up the wrong nodes (no idea why, still need to go look for the PUT log) and some of the primaries 404'd (the frags were on handoffs) and the nodes that do come back seems like they should be enough but the proxy barfs with no log message01:10
claygpeluse: i was just troubleshooting with swifterdarrell how come ssync doesn't work with replication networks!01:11
claygpeluse: turns out that @replication decorator doesn't make any allowances for the pipelined ssync requests - so replication_server = true means you can't use ssync01:11
*** DCWilliams_VA has quit IRC01:12
claygfor now we'll probably just make our replication servers allow all verbs - that works around the 405's in the subrequests anyway01:12
claygneat trick there I could have used the replication_ip for the fragment rebuild GET requests after all!01:12
claygalso, I really need to update probetests to work with seperate replication networks so I can run my dev setup like that sometimes01:13
claygbut last time i tried it looked like it was going to be a ton of work :'(01:13
kota_mattoliverau: was good ;)01:26
*** jrichli has joined #openstack-swift01:26
*** annegentle has joined #openstack-swift01:56
*** erlon has quit IRC02:01
*** annegentle has quit IRC02:02
*** jkugel has joined #openstack-swift02:04
*** wbhuber has joined #openstack-swift02:09
*** shakamunyi has joined #openstack-swift02:09
*** kota_ has quit IRC02:11
*** bkopilov has quit IRC02:35
*** thumpba has joined #openstack-swift02:43
*** jkugel has quit IRC02:49
*** panbalag has quit IRC02:54
*** annegentle has joined #openstack-swift02:57
*** silor has joined #openstack-swift03:01
*** wbhuber has quit IRC03:09
claygindeed python y u no ordered dict literal03:21
*** annegentle has quit IRC03:25
claygok, so segfault issue has been emailed - there's reconstructor bug in there too I think - i guess I'll file that, and the replication network ssync thing, and then maybe back to the GET issue03:41
*** links has joined #openstack-swift03:45
*** tamizh_geek has joined #openstack-swift03:49
*** bkopilov has joined #openstack-swift04:01
*** asettle has joined #openstack-swift04:02
*** kota_ has joined #openstack-swift04:17
*** jrichli has quit IRC04:18
*** tamizh_geek has quit IRC04:23
*** zaitcev has quit IRC04:38
*** SkyRocknRoll has joined #openstack-swift04:58
*** SkyRocknRoll has joined #openstack-swift04:58
*** tamizh_geek has joined #openstack-swift05:11
openstackgerritHisashi Osanai proposed openstack/swift: WIP: Enable Role-based access control using oslo.policy in Swift  https://review.openstack.org/14993005:19
claygbah why is zaitcev never around when you need him - roflmao @ object-supertransmogrificator05:23
mattoliverauYay, killed my cloud server SAIO, turns out upscaling testing is usefult to find problems, who whould have thought! Back to debugging :)05:28
kota_clayg: what's object-supertransmogrificator?05:29
kota_clayg: sounds interesting05:29
claygidk, I *think* zaitcev made it up - it *does* sound interesting, right!  :D05:30
kota_hah05:30
claygnotmyname: charz: are you running benchmarks on the qa cluster05:33
claygnotmyname: charz: I'm finally getting back around to looking at the GET issue05:33
charzclayg: no, did you know where is logs?05:35
claygcharz: yeah I think i found them, I'm confused about ssbench on a read scenario05:49
claygif you do 100% reads - it has to get the objects from somewhere - so it will pre-seed data if it needs to?05:50
claygbut what about if you have a previous run - is there any chance we're not wiping the cluster between benchmarks?05:50
charzclayg: yes, it05:50
claygI don't know how else my greps could be missing this objects PUT unless they were in logs that got rolled out or it was PUT in a different run05:51
charzit will upload objects and get it back05:51
claygso i have this object who's fragments are over the wrong nodes - just out of order mostly - so i want to understand why the PUT went so bad - but I don't see a PUT in the current log set for this object name - just GETs05:52
clayg... and a HEAD05:52
charzclayg: which scenario/log you are looking?05:53
charzclayg: ec_small_test_read_only?05:54
claygyeah i'm pretty sure the pre-seed option is doing a HEAD check before the PUT and these objects are all on the wrong nodes from a previous run06:01
hoI would like to modify https://github.com/openstack-dev/devstack/blob/master/lib/swift to pass check-* for #149930 in gerrit. Is there any rule to modify it?06:04
charzclayg: I think I can do a small change(object name) in scenario file, that'll avoid this problem. And ssbench will put the object to new path.06:08
claygand there you go - Returning 503 for %s; found too many etags (%s) <- debug log level!06:13
claygso there's two fragments with index 0 and one of them is an older timestamp - when it get's picked up in the list of six there's a debug message about hey I'm totally going to throw this request on the floor and barf for this very specific reason - but shhhhhh - don't tell anyone06:14
kota_ho: you mean you want to change the gate?06:14
claygfwiw this acctually *would* get better if we had a working reconstrcutor config on these nodes06:14
kota_ho: if you want to make any changes the gerrit gate, you should push the changes to openstack-infra project.06:15
hokota_: yes. I would like to add a new file to /etc/swift/06:15
hokota_: thanks! but my patch is not approved so is it ok to push the modification?06:16
kota_I didn't see the patch yet06:16
claygpatch 14993006:16
patchbotclayg: https://review.openstack.org/#/c/149930/06:16
kota_ho: maybe... no?06:16
claygisn't that the cart before the horse tho?  should the gate pass with your change without requiring new configs - this is called "backwards compatible"06:17
hoclayg: it is!06:17
openstackgerritPrashanth Pai proposed openstack/swift: Refactor server side copy as middleware  https://review.openstack.org/15692306:17
openstackgerritPrashanth Pai proposed openstack/swift: versioned writes middleware  https://review.openstack.org/13434706:17
kota_clayg: right06:17
kota_ho: if the change for the gate has backward compatibility, we can push it to gerrit always.06:18
kota_ho: i don't know patch 149930 depends on whatever but if the gate change needs 149930, we should push it after 149930 merged, IMO.06:21
patchbotkota_: https://review.openstack.org/#/c/149930/06:21
claygg'night06:21
kota_clayg: good night, have a good sleep ;)06:22
claygnotmyname: I'm pretty sure I know the 503's on GET are from - they are related but subtley different - there's no "fix" except make the GET smarter, which we might have to do, but it would probably have never been noticed except the qa cluster's reconstructor isn't running06:22
claygnotmyname: charz is going to fix the configs and I'll chat with torgomatic more about how we can maybe think about imporving the "throw out frags with crap old etags and look for more before you 503" situation in the proxy06:23
hokota_: clayg: thanks :-) now i'm thinking...06:23
hoclayg: good night!06:23
hokota_: I would like to add a file /etc/swift but this file is not used without my patch. means just push a patch to devstack has backward compatibility right?06:25
kota_ho: sounds it has a compatibility but06:27
kota_ho: it looks unnecessary to devstack users until your pach merged, right?06:28
hokota_: right.06:31
kota_ho: imho, the better way could be make tests w/o the file you want to add (something like using mock) and then06:31
kota_s/make/to make/06:31
kota_ho: add the file the stuff needs to devstack (or infra-ish) after your change landed on master.06:33
kota_ho: and then, create more tests using the real files you added.06:33
openstackgerritShri Javadekar proposed openstack/swift: Create object files directly in datadir.  https://review.openstack.org/18088306:34
hokota_: i see. i will have the content of the file in the code now. thanks for your help/ :-)06:36
kota_ho: nop, we want to keep the tracking files clean as possible and we can add files *when we need it* :P06:38
*** kota_ has quit IRC06:46
*** joeljwright has joined #openstack-swift06:55
*** thumpba has quit IRC07:08
openstackgerritChristian Schwede proposed openstack/swift: Increase replicator log level if nothing got replicated  https://review.openstack.org/18089307:17
*** joeljwright has quit IRC07:23
openstackgerritDaisuke Morita proposed openstack/swift-specs: Changing Policies spec  https://review.openstack.org/16876107:29
*** krykowski has joined #openstack-swift07:31
cschwedenotmyname: I added you as a reviewer for the spec on adding cors in openstack (patch 179866)07:36
patchbotcschwede: https://review.openstack.org/#/c/179866/07:36
cschwedei commented on the spec earlier, and there was a change regarding swift+global cors middleware. pls have a look if you find some time07:37
*** jordanP has joined #openstack-swift07:41
openstackgerritHisashi Osanai proposed openstack/swift: WIP: Enable Role-based access control using oslo.policy in Swift  https://review.openstack.org/14993007:41
*** acoles_away is now known as acoles07:53
*** jistr has joined #openstack-swift07:54
*** dmorita has quit IRC07:59
*** joeljwright has joined #openstack-swift08:00
openstackgerritChristian Schwede proposed openstack/swift-specs: Prevent accidental object deletion spec  https://review.openstack.org/15542108:03
openstackgerritChristian Schwede proposed openstack/swift: Increase replicator log level if nothing got replicated  https://review.openstack.org/18089308:07
*** geaaru has joined #openstack-swift08:18
*** chlong has quit IRC08:23
openstackgerritHisashi Osanai proposed openstack/swift: WIP: Enable Role-based access control using oslo.policy in Swift  https://review.openstack.org/14993008:34
*** winggundamth has joined #openstack-swift09:06
winggundamthhi. I follow http://docs.openstack.org/kilo/install-guide/install/apt/content/swift-install-controller-node.html with keystone but after I try doing swift list I got "Invalid user token - deferring reject downstream" error09:07
winggundamthI can do keystone user-list but not swift-list. This is full log http://paste.openstack.org/show/214954/09:07
*** silor has quit IRC09:11
*** aix has joined #openstack-swift09:14
openstackgerritChristian Schwede proposed openstack/swift-specs: Add event notification spec  https://review.openstack.org/18091409:46
acoleswinggundamth: it looks like the keystone authtoken middleware is failing to validate your token - double check your config options for the authtoken middleware, and (if you can) verify that it is communicating with your keystone service (from its logs)09:49
winggundamthacoles: let me recheck my configuration. I think I already follow everything as docs said09:49
winggundamthacoles: just wonder if the docs right http://docs.openstack.org/kilo/install-guide/install/apt/content/swift-install-controller-node.html in [filter:authtoken] it used auth_uri and auth_url for var name. but on juno it used auth_uri and identity_uri with admin_ as prefix for user/pass/tenant name09:52
acoleswinggundamth: i use the config described here http://docs.openstack.org/developer/swift/overview_auth.html#configuring-swift-to-use-keystone09:57
acolesie identity_uri09:58
acoleswinggundamth: i'm not familiar with the password plugin described in the doc you linked09:58
acoleswinggundamth: btw auth_uri is only ever used to return to a client, not used to validate tokens, so that is not cause of problem09:59
acoleswinggundamth: i'm just going to try the password plugin options...10:00
winggundamthacoles: http://paste.openstack.org/show/216005/ This is my proxy-server configuration10:00
winggundamthacoles: I used self-signed for https10:00
winggundamthacoles: and let me try the one you sent also10:03
winggundamthacoles: when used config the one you've sent. I still got the same error10:07
openstackgerritChristian Schwede proposed openstack/swift-specs: Add metadata search spec  https://review.openstack.org/18091810:08
winggundamthacoles: I see this error in account-server log "HEAD /swift01/53488/AUTH_bebf7840e5914834968571d068ceff24" 404 - "HEAD http://storage.example.com/v1/AUTH_bebf7840e5914834968571d068ceff24" "tx463e99d6057d413aa8abb-00554b3956" "proxy-server 137" 0.0002 "-" 22710:08
acoleswinggundamth: ok, i cannot get the password plugin method to work from doc you linked :(10:13
acoleswinggundamth: so is the account-server log from after you changed your auth token options?10:13
winggundamthacoles: yes after I change to your configuration10:14
winggundamthacoles: just wonder I couldn't make your config working either. Do you think this is Kilo bug?10:14
acoleswinggundamth: so with 'my' config method do you still get a 401 or are you now getting a 404?10:15
acolesbtw i'm running kilo10:15
winggundamthacoles: I still got 401 with your config10:16
winggundamthacoles: http://paste.openstack.org/show/216012/ this is config in authtoken with your config10:16
winggundamthlet me test that user/pass is correct10:17
winggundamthyes confirm that swift user/pass is correct since I can do keystone token-get command10:18
acoleswinggundamth: you have no port on the identity_uri10:18
winggundamthacoles: yes. this is expected since I'm using nginx to forward port to real identity port10:19
winggundamthacoles: and I can confirm that I can accessing to that domain and identity_url correct in endpoint10:21
winggundamthacoles: but I wonder in account-server log that I got 404 "HEAD /swift01/53488/AUTH_bebf7840e5914834968571d068ceff24" 404 - "HEAD http://storage.example.com/v1/AUTH_bebf7840e5914834968571d068ceff24" "tx925d24a0d7274fd68a4d5-00554b3947" "proxy-server 135" 0.0002 "-" 22910:22
winggundamththe url of storage.example.com is http not https10:22
winggundamthacoles: checking with juno one that on our production right now and it's correct that url will be http10:26
acoleswinggundamth: i think the 404 from the account server is normal until you actually create a container in the account - with account_autocreate=true the account doesn't actually exist until you create a container10:26
winggundamthacoles: i see10:26
winggundamthacoles: so do you think the problem is from proxy-server configuration right?10:27
openstackgerritChristian Schwede proposed openstack/swift-specs: Add metadata search spec  https://review.openstack.org/18091810:27
acolesi can see from your paste you do have account_autocreate=true which is correct10:27
*** silor has joined #openstack-swift10:28
acoleswinggundamth: can you paste your proxy log with current auth token config?10:28
winggundamthsure10:28
acoleswinggundamth: and maybe bump your log level to debug10:30
winggundamthacoles: http://paste.openstack.org/show/216013/ this normal one first10:30
winggundamthacoles: this is debug one http://paste.openstack.org/show/216014/10:32
winggundamthacoles: I see it that it authorized as anonymous10:33
*** ho has quit IRC10:36
acoleswinggundamth: if you scroll right on your last paste you can see the token validation failure is due to SSLError: SSL exception connecting to https://admin-identity.example.com10:38
winggundamthacoles: so is it a self-signed bug problem?10:38
winggundamthacoles: should I try to upgrade keystoneclient to latest?10:42
acoleswinggundamth: you can try - auth token is from keystonemiddleware not keystoneclient10:44
acoleswinggundamth: sorry i'm out of ideas to help - other than it appearing to be auth_token not being able to connect to your keystone service10:45
acoleswinggundamth: i'm stepping away for a bit10:46
*** acoles is now known as acoles_away10:47
winggundamthacoles: ok I'm out too. thank you so much for your help10:49
*** cdelatte has quit IRC10:51
*** acoles_away is now known as acoles11:03
*** panbalag has joined #openstack-swift11:07
*** links has quit IRC11:31
*** links has joined #openstack-swift11:32
*** kota_ has joined #openstack-swift11:41
*** ppai has joined #openstack-swift11:50
*** openstackgerrit has quit IRC11:52
*** openstackgerrit has joined #openstack-swift11:52
*** lpabon has joined #openstack-swift12:01
*** thurloat_isgone is now known as thurloat12:03
*** ppai has quit IRC12:11
*** waterkinfe has joined #openstack-swift12:13
*** waterkinfe has quit IRC12:15
*** fthiagogv has joined #openstack-swift12:25
*** cdelatte has joined #openstack-swift12:28
*** proteusguy has joined #openstack-swift12:33
*** proteusguy has quit IRC12:34
*** tamizh_geek has quit IRC12:34
*** tamizh_geek has joined #openstack-swift12:35
*** bkopilov has quit IRC12:39
*** tamizh_geek has quit IRC12:39
*** fthiagogv has quit IRC12:43
*** silor has quit IRC12:44
*** dencaval has joined #openstack-swift12:45
*** fthiagogv has joined #openstack-swift12:46
*** ppai has joined #openstack-swift12:48
winggundamthacoles_away: I got it working :) you have to specify insecure = True in authtoken12:51
*** jkugel has joined #openstack-swift12:54
*** ppai has quit IRC12:55
winggundamthacoles_away: I also test with config as in kilo installation and it works too12:57
openstackgerritChristian Schwede proposed openstack/swift-specs: Add event notification spec  https://review.openstack.org/18091412:58
*** kei_yama has quit IRC13:02
*** NM has joined #openstack-swift13:15
*** jkugel has left #openstack-swift13:16
acoleswinggundamth: great, glad you have it working, i've not used ssl in my SAIO so never set that option13:19
*** david-lyle has quit IRC13:19
acoleswinggundamth: now i am wondering why i could not get the kilo instructions to work for me :/13:19
winggundamthacoles: haha. it's your job now. I'll post mine configuration in case of you want it13:21
acolesheh13:21
winggundamthacoles: http://paste.openstack.org/show/216163/ here you go13:22
acoleswinggundamth: thx13:23
*** openstackstatus has quit IRC13:25
*** openstackstatus has joined #openstack-swift13:26
*** ChanServ sets mode: +v openstackstatus13:26
*** kota_ has quit IRC13:38
*** chlong has joined #openstack-swift13:39
*** SkyRocknRoll has quit IRC13:39
openstackgerritEmmanuel Cazenave proposed openstack/swift: X-Auth-Token should be a byestring.  https://review.openstack.org/18009813:44
*** esker has joined #openstack-swift13:50
*** jamielennox is now known as jamielennox|away13:51
*** fanyaohong has joined #openstack-swift13:54
*** links has quit IRC13:54
*** zul has quit IRC13:57
*** delattec has joined #openstack-swift13:59
*** zul has joined #openstack-swift14:02
*** cdelatte has quit IRC14:02
*** jrichli has joined #openstack-swift14:06
*** Nadeem has joined #openstack-swift14:08
openstackgerritAlistair Coles proposed openstack/swift: Filter Etag key from ssync replication-headers  https://review.openstack.org/17397314:12
acolesclayg: i made this ^^ how i think you'd prefer it14:13
*** atan8 has joined #openstack-swift14:14
openstackgerritAlistair Coles proposed openstack/swift: Filter Etag key from ssync replication-headers  https://review.openstack.org/17397314:18
*** bkopilov has joined #openstack-swift14:20
*** wbhuber has joined #openstack-swift14:21
*** rookie has joined #openstack-swift14:26
*** rookie has quit IRC14:26
*** hdworker has joined #openstack-swift14:27
openstackgerritAlistair Coles proposed openstack/swift: X-Auth-Token should be a bytestring.  https://review.openstack.org/18009814:28
*** hdworker has quit IRC14:35
*** Nadeem has quit IRC14:47
*** minwoob has joined #openstack-swift14:47
*** tamizh_geek has joined #openstack-swift14:48
*** nadeem has joined #openstack-swift15:08
notmynamegood morning15:09
notmynameclayg: thanks for looking more at the read errors15:09
*** annegentle has joined #openstack-swift15:09
notmynamecschwede: thanks15:10
notmynamecschwede: although I think it's _really_ strange to say "hey let's just add CORS support to eg a service that associates block devices with VMs". ie CORS is for browsers. it doesn't make much sense to implement it for most projects15:12
*** annegentle has quit IRC15:14
*** annegentle has joined #openstack-swift15:14
openstackgerritMerged openstack/swift: Add missing statsd metrics section for object-reconstructor  https://review.openstack.org/18065515:33
*** krykowski has quit IRC15:38
*** fthiagogv_ has joined #openstack-swift15:46
*** fthiagogv has quit IRC15:48
*** fthiagogv_ has quit IRC15:50
*** fthiagogv has joined #openstack-swift15:50
*** zaitcev has joined #openstack-swift15:54
*** ChanServ sets mode: +v zaitcev15:54
*** fanyaohong has quit IRC15:59
notmynamejrichli: good morning16:03
*** bkopilov has quit IRC16:04
*** nadeem has quit IRC16:04
*** hdworker has joined #openstack-swift16:05
jrichligood morning.  what's up?16:05
notmynamejrichli: pm sent16:06
*** bkopilov has joined #openstack-swift16:09
*** winggundamth has quit IRC16:15
*** madgriff has joined #openstack-swift16:28
*** shakamunyi has quit IRC16:29
notmynameI have to plug in to my big external monitor to do summit calendar stuff. my laptop screen is too big to track it all16:37
*** atan8 has quit IRC16:38
*** hdworker has quit IRC16:38
*** annegentle has quit IRC16:39
*** cazino has left #openstack-swift16:48
*** jordanP has quit IRC17:02
*** tamizh_geek has quit IRC17:04
*** setmason has joined #openstack-swift17:09
setmasonAnybody know of a swift version of apt-transport-s3 (https://github.com/kyleshank/apt-transport-s3)? Which would allow you to store your apt repo directly in swift?17:12
*** zhill has joined #openstack-swift17:16
*** jistr has quit IRC17:17
*** thumpba has joined #openstack-swift17:21
*** thumpba_ has joined #openstack-swift17:22
*** thumpba has quit IRC17:26
*** tamizh_geek has joined #openstack-swift17:38
*** tacotuesday has joined #openstack-swift17:39
*** annegentle has joined #openstack-swift17:40
*** madgriff has quit IRC17:42
*** erlon has joined #openstack-swift17:43
*** MVenesio has joined #openstack-swift17:45
*** anteaya has quit IRC17:45
*** bejorg has joined #openstack-swift17:45
*** annegentle has quit IRC17:46
bejorgHi guys, question regarding Swift quarantine zone, is there is way to see content ? Do we have to clear this quarantine zone ?17:46
bejorgI have some objects inside as displayed by swift-recon17:46
*** vinsh has joined #openstack-swift17:47
*** jistr has joined #openstack-swift17:49
*** nadeem has joined #openstack-swift17:50
*** theanalyst has quit IRC17:52
*** nadeem has quit IRC17:54
*** geaaru has quit IRC17:55
*** theanalyst has joined #openstack-swift17:55
*** acoles is now known as acoles_away17:56
*** SkyRocknRoll has joined #openstack-swift18:02
*** SkyRocknRoll has quit IRC18:03
*** SkyRocknRoll has joined #openstack-swift18:03
*** SkyRocknRoll has quit IRC18:03
*** SkyRocknRoll has joined #openstack-swift18:03
*** tellesnobrega has quit IRC18:05
*** annegentle has joined #openstack-swift18:06
*** nadeem has joined #openstack-swift18:07
*** dencaval has quit IRC18:10
*** jrichli has quit IRC18:12
*** jordanP has joined #openstack-swift18:12
*** jordanP has quit IRC18:13
*** tacotuesday has quit IRC18:14
*** jrichli has joined #openstack-swift18:18
*** tellesnobrega has joined #openstack-swift18:22
*** SkyRocknRoll has quit IRC18:32
zaitcevbejorg: As far as I know, it is necessary to clean up the quarantine area manually. In practice, however, Swift rarely exhibits bugs that quarantine improperly. So, if it starts filling up, your filesystem and/or device is toast and the node needs to be replaced.18:36
zaitcevbejorg: I am not an operator, however. Maybe a good question to ask at the openstack-operators list.18:36
bejorg I will, thanks for informations18:36
*** tellesnobrega has quit IRC18:41
*** nadeem has quit IRC18:43
*** tamizh_g_ has joined #openstack-swift18:47
*** tamizh_geek has quit IRC18:49
*** SkyRocknRoll has joined #openstack-swift18:49
*** jistr has quit IRC18:51
*** nadeem has joined #openstack-swift18:53
*** dencaval has joined #openstack-swift18:54
claygmorning18:58
*** nadeem has quit IRC19:05
*** nadeem has joined #openstack-swift19:06
*** nadeem has quit IRC19:07
*** fthiagogv has quit IRC19:09
*** zul has quit IRC19:16
openstackgerritPete Zaitcev proposed openstack/swift: Spell "rebalance" right in swift-ring-builder.1  https://review.openstack.org/18112219:17
zaitcevI'm happy when people file bugs. I really am.19:18
zaitcevHeh.19:18
notmynamezaitcev: I like how gerrit auto-links that bug to launchpad and shows something for ubuntu19:20
*** tellesnobrega has joined #openstack-swift19:20
notmynamezaitcev: however, if you're tracked by how many tickets you close, this is great ;-) https://bugzilla.redhat.com/show_bug.cgi?id=121826919:21
openstackbugzilla.redhat.com bug 1218269 in openstack-swift "Action rebalence should be rebalance for command swift-ring-builder" [Unspecified,New] - Assigned to zaitcev19:21
zaitcevnotmyname: It's worse. When I checked out, I did  git checkout -b rhbz/1218269  (because I knew not to use bug/1218269),  but it renamed the topic into "bug/1218269".19:21
claygwat?19:24
*** tamizh_geek has joined #openstack-swift19:24
claygheh, nice19:24
notmynamezaitcev: you should follow torgomatic's example and make branches named with funny puns19:25
claygbejorg: that's intresting!  What's in the objects?  Are they fully replicated and durable in the correct parts?  How old are they?  Can you read metadata off them!?19:25
*** tamizh_g_ has quit IRC19:26
zaitcevclayg: Before bejorg can answer that, he has to find where the objects are placed and how to read the metadata (well, I can imagine running pyton, import pickle, p=pickle.load(open(xxxxx)) etc.)19:27
claygsetmason: hey that looks sweet!  You could try it out with swift + s3 - it looks like it uses the domain/bucketname syntax instead of bucketname.domain - so it might work - or you could file an issue with the maintainer to add swift support19:27
claygzaitcev: yeah swift-object-info will tell you most of that if the metadata is still there19:28
zaitcevoh, it takes a full filename19:28
claygif not it's more fun!  you have the hash (or you might, sometimes I think other things will get quarantined - depends what happened!)  very exciting19:28
*** zul has joined #openstack-swift19:31
claygnotmyname: so, are we doing better in the qa cluster post-fix-the-reconstructor19:32
notmynameclayg: yay19:33
claygyay or yeah?19:33
notmynamethanks for looking in to that. I've been swamped with summit talk prep this morning19:33
claygoh - bummer19:33
notmynameoh, I saw "we are doing better..."19:33
*** bkopilov has quit IRC19:33
notmynameI haven't looked19:33
claygheh - k19:33
*** SkyRocknRoll has quit IRC19:33
notmynamecharz has some results, they looked ok-ish, but he cranked up concurrency again19:34
claygcrank it up!19:34
claygtorgomatic: so - broken unittest one the PUT fix for quorum?19:34
claygor... broken code?19:34
*** bkopilov has joined #openstack-swift19:43
bejorgclayg: zaitcev: no idea what's inside objects. I just saw quarantine from swift-recon. From my understanding, auditor process moved them into quarantine zone (where is it ? do we have to flush it ?), and them replicator process has copied from health replica19:46
bejorgjust 17 quarantine objects... curious to know if the swift-recon stats will go back to 0 or not...19:47
notmynamebejorg: there's a top-level directory in each device named "quarantine". that's where it's stored19:47
bejorgok ! checking19:48
notmynamebejorg: and you can use swift-object-info to inspect the individual .data files there19:48
notmynamebejorg: and, no, that's not automatically cleaned up19:49
bejorgok but how do I know "failed" objects ? if you have 100000 objects...19:49
bejorgnice, I got them from the quarantined folder19:50
*** bkopilov has quit IRC19:53
*** bkopilov has joined #openstack-swift19:53
zaitcevObjects are all self-identified, so swift-object-info may be used to parse the xattrs if available. However, it's usually not needed to try and "restore" them. The files may have bad sectors or whatnot. So what you need is track if excessive quarantining going on at the node, like you did.19:54
*** tamizh_geek has quit IRC19:55
bejorgthen, it is safe to delete quarantined folder content ? rm -f /path/to/quarantined/folder/* ?19:55
*** thumpba_ has quit IRC19:56
zaitcevno19:57
zaitcevWell, at least "not without testing". Like I said, what if those files cover bad blocks? If you delete them, you release those blocks into free pool.19:58
bejorgok, then how can I clear each quarantined folder ? right now, that pretty small on my case. 17 objects on quarantined = ~100MB19:58
bejorggood point!19:58
zaitcevI'd cat them into /dev/null at least, see if it works19:59
*** bkopilov has quit IRC19:59
zaitcevonly then rm per your message above19:59
*** bkopilov has joined #openstack-swift19:59
*** delattec has quit IRC19:59
*** annegentle has quit IRC20:02
*** annegentle has joined #openstack-swift20:02
bejorgcat object > /dev/null all passed. what should I see in case of bad block ?20:03
*** dencaval has quit IRC20:03
zaitcevEIO from cat20:05
zaitcevI think you should be fine to delete them, but it's curious why they fell into quarantine. Maybe we have a bug that makes back-ends corrupt checksums, forget to store checksums, or some other misbehavior and your cluster makes it show up.20:06
zaitcevExcessive crashes could do that20:06
bejorga swift-object-info on any quarantined objects says : Etag: 2af48fde1100d2cbc3d621c5ff84898f doesn't match file hash of 5ab66c6cd4eb3f876774473831787002!20:09
bejorgchecksum were not equal20:09
*** jkremer has joined #openstack-swift20:14
*** breitz has quit IRC20:17
*** breitz has joined #openstack-swift20:17
*** thurloat is now known as thurloat_isgone20:31
claygoh interesting - so bit rot maybe20:32
claygbejorg: you can turn off the automatic checksum in swift-object-info (there's some flag, try --help, or calling it with no args) - might go ahead and print out the object info if it's not already20:32
bejorgswift-object-info -n20:47
*** lpabon has quit IRC20:51
notmynametorgomatic: you just really needed to get your review stats up? ;-) https://review.openstack.org/#/c/181122/20:51
torgomaticnotmyname: the exchange rate between Gerrit review stats and Reddit karma is looking especially good these days ;)20:52
notmynameI have you know that I once submitted an article, as an Anonymous Coward, to Slashdot and it made it to the front page. that's worth at least *some* reddit karma.20:53
torgomaticit's the broker fees that get you20:57
*** annegentle has quit IRC21:04
*** Anticime1 is now known as Anticimex21:14
*** EmilienM is now known as EmilienM|afk21:16
openstackgerritJanie Richling proposed openstack/swift: WIP - working on the encryption feature.  https://review.openstack.org/15790721:29
*** jrichli has quit IRC21:30
*** MVenesio has quit IRC21:37
*** jkremer has quit IRC21:37
openstackgerritMerged openstack/swift: Spell "rebalance" right in swift-ring-builder.1  https://review.openstack.org/18112221:40
*** NM has quit IRC21:45
*** bill_az has joined #openstack-swift21:47
*** david-lyle has joined #openstack-swift21:56
*** chlong has quit IRC21:59
*** david-lyle has quit IRC22:38
*** esker has quit IRC22:42
*** ujjain has quit IRC22:48
*** km has joined #openstack-swift22:51
*** ujjain has joined #openstack-swift22:52
*** ujjain has joined #openstack-swift22:52
*** vinsh has quit IRC23:00
mattoliverauMorning all23:02
torgomatichi23:02
*** wbhuber has quit IRC23:13
notmynameclayg: I'm going to look at the qa cluster again. have you done anything on it today?23:17
claygnotmyname: NOPE23:19
notmynameok23:19
notmynamevery few errors. still a lot of retires. didn't charz say those were IOError things?23:20
notmynameI'd like to rerun with a low concurrency23:21
notmynamesee if those retries can be eliminated23:21
notmynameclayg: what should I be looking at about the weird placement of the FAs? did you find anything there last night?23:23
claygnotmyname: they were pre-existing objects23:23
claygcharz said that he fixed the ssbench configs so every run will use new object names (even if they're read only)23:23
notmynameah23:24
*** chlong has joined #openstack-swift23:24
notmynameok, I'll verify that. thanks23:24
claygnotmyname: anyway the pre-existing objects were in the wrong place because god only knows what you guys did to those rings - and there was no rebalancer/reconstructor running23:24
notmynameyup23:24
notmynameI thought each run was using a new container name23:24
notmynamethat's what I get for thinkin'23:24
claygnotmyname: either way it'd be boss to verify the reconstructor is working now - and that if there is some data in the cluster it mostly seems in the right place23:24
claygi could check that23:24
claygnotmyname: well I'd look at a benchmark run for a read job - and wehre you'd expect a bunch of PUTs in the logs all I would find is a HEAD23:25
claygi should have spot checked the timestamps to see if it jived with my assersion they existed before the benchmark started - but the logs seemed to support it - I never tracked down the PUTs23:25
clayg... and knowing that reconstruction couldn't rebalance - i guess I considered my theory validated :\23:26
notmynameok23:27
notmynameI don't see a reconstructer running on 0123:27
notmynamehmm.. there is a config section in 2.conf23:28
claygnotmyname: bummer23:28
*** kei_yama has joined #openstack-swift23:34
*** vinsh has joined #openstack-swift23:38
*** ho has joined #openstack-swift23:46
*** kota_ has joined #openstack-swift23:51
*** rmcall has joined #openstack-swift23:53
*** minwoob has quit IRC23:55

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