Thursday, 2014-07-17

*** pberis has quit IRC00:02
*** pberis has joined #openstack-swift00:04
*** kevinc_ has quit IRC00:06
*** pberis has quit IRC00:13
*** pberis has joined #openstack-swift00:16
*** bszabo has joined #openstack-swift00:25
*** bszabo has quit IRC00:30
*** shri has left #openstack-swift00:38
*** pberis has quit IRC00:40
*** pberis has joined #openstack-swift00:42
*** pberis has quit IRC00:51
*** mwstorer has quit IRC00:52
*** pberis has joined #openstack-swift00:53
*** jthorne has joined #openstack-swift00:57
openstackgerritA change was merged to openstack/swift: Disable case-changing behavior in Eventlet
*** foexle_ has joined #openstack-swift01:06
*** foexle has quit IRC01:10
*** pberis has quit IRC01:43
*** pberis has joined #openstack-swift01:44
*** nosnos has joined #openstack-swift01:49
*** haomaiw__ has joined #openstack-swift02:04
*** haomaiwang has quit IRC02:07
*** pberis has quit IRC02:21
*** pberis has joined #openstack-swift02:23
*** bkopilov has quit IRC02:37
*** pberis has quit IRC02:43
*** pberis has joined #openstack-swift02:45
*** tongli has joined #openstack-swift02:49
*** chandankumar has joined #openstack-swift02:51
*** chandankumar has quit IRC02:52
*** chandankumar has joined #openstack-swift02:55
*** nosnos has quit IRC03:06
*** tong_ has joined #openstack-swift03:24
*** tong_ has quit IRC03:24
*** tongli has quit IRC03:24
*** tongli has joined #openstack-swift03:25
*** nosnos has joined #openstack-swift03:33
*** pberis has quit IRC03:34
*** chandankumar has quit IRC03:34
*** pberis has joined #openstack-swift03:36
*** tongli has quit IRC03:38
*** tongli has joined #openstack-swift03:39
openstackgerritYuan Zhou proposed a change to openstack/python-swiftclient: Allow to specify storage policy when uploading objects
*** chandankumar has joined #openstack-swift03:48
*** tongli has quit IRC03:59
*** bkopilov has joined #openstack-swift04:00
*** bszabo has joined #openstack-swift04:02
*** bszabo has quit IRC04:06
*** zaitcev has quit IRC04:19
*** judd7 has quit IRC04:44
*** chandankumar has quit IRC04:48
*** ppai has joined #openstack-swift04:53
*** chandankumar has joined #openstack-swift04:56
*** chandankumar has quit IRC04:57
*** anand_ts has joined #openstack-swift04:58
*** chandankumar has joined #openstack-swift04:59
*** psharma has joined #openstack-swift05:12
*** kota_ has joined #openstack-swift05:19
*** kota_ has quit IRC06:00
*** mmcardle has joined #openstack-swift06:08
*** k4n0 has joined #openstack-swift06:22
*** foexle_ has quit IRC06:29
*** patchbot` has joined #openstack-swift06:30
*** patchbot has quit IRC06:30
*** patchbot` is now known as patchbot06:32
*** dosaboy has joined #openstack-swift06:33
*** ahale_ has joined #openstack-swift06:33
*** pberis has quit IRC06:34
*** amandap has quit IRC06:34
*** fleducquede_ has quit IRC06:34
*** psharma has quit IRC06:34
*** rando has quit IRC06:34
*** d0ugal has quit IRC06:34
*** zackmdavis has quit IRC06:34
*** Tyger has quit IRC06:34
*** clayg_ has joined #openstack-swift06:34
*** mdonohoe has joined #openstack-swift06:34
*** JelleB has quit IRC06:34
*** omame has quit IRC06:34
*** dosaboy_ has quit IRC06:34
*** IRTermite has quit IRC06:34
*** k4n0 has quit IRC06:34
*** ahale has quit IRC06:34
*** pconstantine has quit IRC06:34
*** clayg has quit IRC06:34
*** pberis has joined #openstack-swift06:34
*** amandap has joined #openstack-swift06:34
*** fleducquede_ has joined #openstack-swift06:34
*** pconstantine has joined #openstack-swift06:35
*** zackmdavis has joined #openstack-swift06:35
*** psharma has joined #openstack-swift06:35
*** omame has joined #openstack-swift06:35
*** JelleB has joined #openstack-swift06:35
*** d0ugal has joined #openstack-swift06:36
*** k4n0 has joined #openstack-swift06:36
*** IRTermite has joined #openstack-swift06:36
*** Tyger has joined #openstack-swift06:36
*** foexle has joined #openstack-swift06:44
*** madhuri has joined #openstack-swift06:54
madhuriHi, is there any documentation available for all the request headers used in swift?06:55
*** msahin_ has quit IRC07:00
*** gvernik has joined #openstack-swift07:18
*** bszabo has joined #openstack-swift07:38
*** bszabo has quit IRC07:43
*** gvernik has quit IRC07:47
*** nacim has joined #openstack-swift07:56
*** zslot has joined #openstack-swift07:58
*** zslot1 has joined #openstack-swift08:01
*** zslot has quit IRC08:03
*** haomaiw__ has quit IRC08:05
*** haomaiwang has joined #openstack-swift08:06
*** mmcardle has quit IRC08:09
*** mmcardle has joined #openstack-swift08:14
*** acoles_away is now known as acoles08:17
*** haomai___ has joined #openstack-swift08:19
*** haomaiwang has quit IRC08:22
*** jthorne has quit IRC08:23
*** ppai has quit IRC08:29
*** mrmoje has joined #openstack-swift08:34
*** ppai has joined #openstack-swift08:35
*** bvandenh has joined #openstack-swift08:35
*** haomai___ has quit IRC08:45
*** haomaiwang has joined #openstack-swift08:45
*** Dev_Jin has joined #openstack-swift08:48
Dev_JinI am trying to figure out if swift storage can be used as local package repository (like say local repository for apt) .. but no luck till now.. any help would be appreciated08:49
*** mkerrin has quit IRC09:16
*** mkerrin has joined #openstack-swift09:23
*** kopparam has joined #openstack-swift09:33
*** bvandenh has quit IRC09:35
*** bszabo_ has joined #openstack-swift09:36
*** kopparam has quit IRC09:40
*** kopparam has joined #openstack-swift09:40
*** Dafna has quit IRC09:43
*** kopparam has quit IRC09:45
*** bvandenh has joined #openstack-swift09:47
*** nosnos has quit IRC09:54
pconstantineDev_Jin: it's simple09:55
pconstantineDev_Jin: just use reprepro to build local apt repo09:55
pconstantineDev_Jin: then upload it as-is to some Swift container09:56
pconstantineDev_Jin: make that container public-readable09:56
pconstantineDev_Jin: that's it09:56
pconstantineDev_Jin: ah, one more thing: do not use https, apt-get is very bad with https09:57
*** ppai has quit IRC09:59
*** Shivani has joined #openstack-swift10:00
*** nosnos has joined #openstack-swift10:01
*** haomaiwang has quit IRC10:08
*** haomaiwang has joined #openstack-swift10:09
*** nosnos has quit IRC10:10
*** haomaiw__ has joined #openstack-swift10:11
*** nosnos has joined #openstack-swift10:12
*** haomaiwang has quit IRC10:14
*** ppai has joined #openstack-swift10:15
Dev_Jinthanks pconstantine.. let me try that out10:23
*** chandankumar has quit IRC10:25
*** chandankumar has joined #openstack-swift10:33
*** nosnos has quit IRC10:38
*** bszabo_ has quit IRC10:47
*** bszabo has joined #openstack-swift10:47
openstackgerritAlistair Coles proposed a change to openstack/swift: Add v2 API to list endpoints middleware
*** bszabo has quit IRC10:52
openstackgerritAlistair Coles proposed a change to openstack/swift: Update doc for list_endpoints v2 API
*** bszabo_ has joined #openstack-swift10:57
acolesclayg: ^^ added this as follow-on patch in case you don't like it10:58
*** ppai has quit IRC10:58
*** ppai has joined #openstack-swift11:01
*** Dafna has joined #openstack-swift11:01
*** pconstantine has quit IRC11:19
*** bszabo_ has quit IRC11:22
*** Dafna has quit IRC11:29
*** mmcardle has quit IRC11:43
*** Dafna has joined #openstack-swift11:43
*** andyandy has joined #openstack-swift11:46
*** mahati has quit IRC11:52
*** bkopilov has quit IRC12:05
*** mmcardle has joined #openstack-swift12:15
*** bkopilov has joined #openstack-swift12:18
*** tongli has joined #openstack-swift12:19
acolesclayg: re replicate & .meta cleanup - yes there is an opportunity to merge .meta's during replication. at the cost of reading the files. maybe it's only done *then* when >N meta's have accumulated (normally merge would be done during a POST).12:33
*** bkopilov has quit IRC12:39
*** miqui has joined #openstack-swift12:41
*** chandankumar has quit IRC12:50
*** pconstantine has joined #openstack-swift12:52
mattoliverauHey all12:57
mattoliverauclayg_, acoles, notmyname, torgomatic: and anyone I missed, could someone take a look at my formpost change ( The infra guys here are looking at uploading test results to swift, and this is a feature they'd love to use.12:59
*** bvandenh has quit IRC12:59
*** k4n0 has quit IRC13:01
*** chandankumar has joined #openstack-swift13:07
*** psharma has quit IRC13:09
*** ppai has quit IRC13:17
*** nacim has quit IRC13:26
*** anand_ts has quit IRC13:28
*** ppai has joined #openstack-swift13:31
*** andyandy has quit IRC13:35
*** kenhui has joined #openstack-swift13:42
*** mahati_ has joined #openstack-swift13:47
*** Manish has joined #openstack-swift13:55
ManishHi All13:55
ManishI have one query.... can any one please let me know that why in SWIFT API documentation, "Create Account" feature is not mentioned, although in SWIFT code we have PUT Account support?13:56
Manishis it because, so that user can only create account in SWIFT (after creating in Keystone) , while creating the first container in that account?13:57
Manishit will be great help, if some one can explain this13:58
pconstantineManish: I suppose it's because creating the account was an "admin only" feature, so it's technically not a part of public API but a part of admin API14:03
*** mrmoje has quit IRC14:04
*** mrmoje has joined #openstack-swift14:04
*** ppai has quit IRC14:11
*** dmsimard_away is now known as dmsimard14:12
*** judd7 has joined #openstack-swift14:15
*** mmcardle has quit IRC14:22
*** Dev_Jin has quit IRC14:22
Manishpconstantine: probably you are right..Thanks :)14:29
*** CybergeekDK has quit IRC14:31
*** mrmoje has quit IRC14:36
*** nacim has joined #openstack-swift14:36
*** mrmoje has joined #openstack-swift14:37
*** CybergeekDK has joined #openstack-swift14:42
*** rando has joined #openstack-swift14:44
*** mmcardle has joined #openstack-swift14:44
*** mdonohoe has quit IRC14:47
*** kevinc_ has joined #openstack-swift14:49
*** haomaiwa_ has joined #openstack-swift14:55
*** pberis1 has joined #openstack-swift14:58
*** clayg has joined #openstack-swift14:58
*** foexle has quit IRC15:01
*** haomaiw__ has quit IRC15:02
*** pberis has quit IRC15:02
*** clayg_ has quit IRC15:02
*** mrmoje has quit IRC15:02
*** Dafna has quit IRC15:02
*** JelleB has quit IRC15:02
*** omame has quit IRC15:02
*** fleducquede_ has quit IRC15:02
*** bkopilov has joined #openstack-swift15:02
*** mrmoje has joined #openstack-swift15:04
*** JelleB has joined #openstack-swift15:06
*** fleducquede_ has joined #openstack-swift15:06
*** Dafna has joined #openstack-swift15:07
*** omame has joined #openstack-swift15:07
acolesclayg: torgomatic: i _almost_ have something (very crude) working, merging .meta's during POST and replication.15:13
*** acoles is now known as acoles_away15:14
torgomaticI wonder why the infra folks are looking at formpost instead of using plain tempurls15:17
*** tdasilva has joined #openstack-swift15:29
peluse_torgomatic:  I've been brainstorming the EC object name collision question for a bit and am coming up with a light drizzle at best... can you restate for me the specific steps/scenario that lead us into the problem situation just to make sure I'm understanding the problem correctly?15:32
*** ujjain has quit IRC15:53
*** kevinc_ has quit IRC15:59
*** chandankumar has quit IRC16:06
*** dencaval has quit IRC16:08
*** tongli has quit IRC16:08
*** tongli has joined #openstack-swift16:09
*** kevinc_ has joined #openstack-swift16:09
*** pberis1 has quit IRC16:11
*** Manish has quit IRC16:12
*** dmsimard_away has joined #openstack-swift16:17
*** mkerrin has quit IRC16:24
*** dmsimard_away has quit IRC16:24
*** mwstorer has joined #openstack-swift16:26
*** pberis has joined #openstack-swift16:32
openstackgerritpaul luse proposed a change to openstack/swift: Initial Erasure Code Docs
*** pberis has quit IRC16:36
*** pberis has joined #openstack-swift16:37
*** pberis has quit IRC16:42
*** zslot1 has quit IRC17:04
notmynamegood morning17:07
*** zaitcev has joined #openstack-swift17:14
*** ChanServ sets mode: +v zaitcev17:14
peluse_good morning!17:15
*** Dafna has quit IRC17:16
*** kevinc_ has quit IRC17:17
notmynamepeluse_: I'm looking over the EC docs right now17:25
notmynamepeluse_: I completely assume it's not a "done" document, but if it's all ok, I'll merge it17:26
notmyname"ok" meaning "what's talked about is reasonable and correct"17:26
*** mmcardle has quit IRC17:27
*** kenhui has quit IRC17:28
*** bvandenh has joined #openstack-swift17:28
*** erlon has quit IRC17:39
notmynamepeluse_: a couple of things on the docs to fix. if you don't get to it soon, I'll try to later today or tomorrow17:43
*** kevinc_ has joined #openstack-swift17:52
*** jthorne has joined #openstack-swift18:24
mattoliverautorgomatic: the infra guys are using form post because they want the test nodes to upload the logs.. And because they could be running arbitory code, and so considered untrusted. So formpost. Thanks for the suggestions I'll implement them when I get in front of a computer :)18:34
torgomaticmattoliverau: yeah, I was thinking why not use tempurl for exactly the same thing, but I think I figured it out18:35
torgomaticwith formpost, you can have the testrunner inject a single thing(swift-url + form params) and upload any arbitrary set of logs you want, so that decouples things better18:37
JoshNanghey guys, can i get reviews on adding tempurls? we've got two drivers in ironic target for juno blocked on it. thanks!!
mattoliverauYeah that's infras point of view too, so soon all the logs being shared through zuul will be in swift ;)18:40
*** bvandenh has quit IRC18:46
*** fifieldt has joined #openstack-swift19:00
*** mmcardle has joined #openstack-swift19:02
notmynameJoshNang: done19:10
notmynameJoshNang: also, nice feature :-)19:10
JoshNangnotmyname: thanks! woo!!19:11
*** kevinc_ has quit IRC19:11
JoshNangnotmyname: any idea when it would get packaged and put in a new release? this is my first swift patch19:12
JoshNangthis patch will use the tempurls in ironic: but from what i understand, i'll have to get global-requirements bumped up after tempurls are packaged19:14
peluse_notmyname:  feel free to take a shot at rewording what I was trying to explain in that paragraph in the EC docs... appreciate it19:15
peluse_torgomatic:  did you see my question earlier?19:17
*** pberis has joined #openstack-swift19:22
*** fifieldt has quit IRC19:31
*** mmcardle has quit IRC19:33
torgomaticpeluse_: no19:42
notmynameJoshNang: well, soon, actually. I want to see land and then cut a release of the client19:42
peluse_torgomatic:   I've been brainstorming the EC object name collision question for a bit and am coming up with a light drizzle at best... can you restate for me the specific steps/scenario that lead us into the problem situation just to make sure I'm understanding the problem correctly?19:43
JoshNangnotmyname: awesome. thanks for your help!19:43
torgomaticpeluse_: so with replication, the replicators will move data closer to home, even if only partway19:43
torgomaticso it might go from 3rd handoff -> 1st handoff, then the 3rd will clean up19:44
torgomaticif you do this with EC stuff, it's harder to tell if the closer handoff really has a copy of *your* fragment archive or if it's got a different one for the same object19:44
*** nacim has quit IRC19:45
peluse_torgomatic:  OK, thanks.  Will explore a few ideas a bit deeper and get back to ya19:46
notmynamepeluse_: for the docs, we'll get it all good before final release. all I'm "worried" about now is making sure we have the right placeholders and the build warning19:47
peluse_good call on the handoffs too, forgot a todo section for htat...19:50
peluse_notmyname / torgomatic:  did you peruse the EC spec yet?19:50
*** mahati_ has quit IRC19:52
*** foexle has joined #openstack-swift19:53
peluse_check it out :)19:55
*** mmcardle has joined #openstack-swift20:00
*** mmcardle has quit IRC20:12
*** kevinc___ has joined #openstack-swift20:13
notmynamepeluse_: still around?20:35
notmynamepeluse_: I want to chat about something that I thought of while reading the EC doc.20:46
notmynamepeluse_: I think there may be a way to improve master today that will also be necessary for EC20:47
notmynamepeluse_: so with EC, the proxy will need to open up concurrent connections to many storage nodes to read an object (object GET)20:48
notmynamepeluse_: today with replication, we try one object server, then another, then another, etc until the data is read20:48
notmynamebut there's been the idea floating around for a while of opening up connections to all the primary nodes and then using the first that responds and dropping the others on the floor20:49
*** fifieldt has joined #openstack-swift20:49
notmynamepeluse_: so an interesting proposal to master would be to add this ability to the object read path, and then reuse it for the EC read path20:50
peluse_notmyname:  I like it20:52
peluse_notmyname:  I'll throw a card up on Trello20:52
notmynamepeluse_: I think it starts with rewriting the _get_source_and_node() method in proxy/controllers/ (line 746)20:52
notmynamepeluse_: IIRC, the arguments against doing this is that it means there will be a lot more sockets (connections) opened in the cluster as a whole20:53
notmynamebut the upside for today's replicated clusters is that we could probably lower time-to-first-byte latency20:53
peluse_notmyname:  yeah, but maybe that socket buffer pool idea can help mitigate the overhead a bit there20:53
peluse_was reading about another obj storage system that used that scheme for PUT also and then used background tasks to rebalance the uneven distrubtion that resulted...20:54
peluse_can't remember where though...20:54
notmynameso anyway, as I was reading through the EC doc, it sounded like that's going to be a crucial part of the EC read path anyway, so I thought it might be nice to solve the general case and get it to master before feature/ec merges20:54
peluse_I put a card on Trello and if nobody else nabs it I'll jump on it after I get some more progress on the obj name collision thing as its blocking me on the reconstructor framework20:55
notmynameawesome. thanks20:55
peluse_no prob, cool idea!20:56
peluse_I pasted our dicsussion in the Trello card :)20:56
*** occupant has quit IRC20:57
*** occupant has joined #openstack-swift20:57
notmynamepeluse_: I think torgomatic has some good ideas on this too (we chatted a little on it in the office). so definitely get his input as you tackle it :-)20:57
peluse_awesome, will do.  need to learn a little more on my own, maybe tomorrow I'll ping20:59
notmynameheh, he's deep in django-landia for now. at some point, he'll emerge with strange and terrible stories of wonder and adventure21:00
peluse_that sounds like loads of fun!21:01
*** mrmoje has quit IRC21:12
*** jthorne has quit IRC21:16
*** tongli has quit IRC21:19
*** occup4nt has joined #openstack-swift21:21
*** occupant has quit IRC21:24
*** occup4nt is now known as occupant21:27
*** occupant has quit IRC21:42
torgomaticwhat are all these new dsvm-blah-havana and -icehouse jobs?21:42
notmynametorgomatic: hmm...look at that. new tests21:43
notmynameno idea21:43
torgomaticwell, it's a good thing that all the tests are 100% reliable and this will in no way make it harder to land changes in Swift21:44
notmynamegate tests. the electric third rail of openstack politics21:44
notmynametorgomatic: like the one that just failed for swiftclient? ;-)21:45
torgomaticnotmyname: something like that, yes :)21:45
*** occupant has joined #openstack-swift21:45
notmynameoh. the test "test_cinder_extra_specs_list" failed. good thing we check that21:47
*** ChanServ sets mode: +v clayg21:56
*** foexle has quit IRC22:15
*** judd7 has quit IRC22:23
MooingLemuris there a way to specify write_affinity to a zone/region which entirely consists of weight-0 devices and have it only use those for incoming writes?  The use case is being able to place ingest nodes at remote locations that can take PUTs at high speed, and have replication push everything into place, but without caring about immediate availability of objects.22:38
*** dmsimard is now known as dmsimard_away22:45
*** kevinc___ has quit IRC22:45
torgomaticMooingLemur: nope; a 0-weight device will have no partitions assigned, so it'll never get any data on it22:46
MooingLemurtorgomatic: Can you think of a way of doing what I want to achieve?  I want storage devices with no partitions actually assigned, but I want them to be used for PUTs, and have replication take over.  I suppose a separate ring known to the proxy server with only the devices I want to write to, but that will break GETs if they happen to come through that proxy.22:48
torgomaticMooingLemur: I think you're out of luck; "no partitions actually assigned" and "used for PUTs" are mutually incompatible the way Swift stands today22:49
*** kevinc_ has joined #openstack-swift22:49
*** elambert has quit IRC22:50
*** Shivani has quit IRC22:50
*** madhuri has quit IRC22:50
*** Shivani has joined #openstack-swift22:51
*** madhuri has joined #openstack-swift22:51
*** kevinc_ has quit IRC22:51
*** clayg has quit IRC22:51
*** ekarlso has quit IRC22:51
*** dfg has quit IRC22:51
*** jokke_ has quit IRC22:51
*** redbo has quit IRC22:51
*** thurloat has quit IRC22:51
*** zaitcev has quit IRC22:52
*** fleducquede_ has quit IRC22:52
*** haomaiwa_ has quit IRC22:52
*** pconstantine has quit IRC22:52
*** otoolee has quit IRC22:52
*** rpedde has quit IRC22:52
*** cschwede has quit IRC22:52
*** swat30 has quit IRC22:52
*** acoles_away has quit IRC22:52
*** sandywalsh has quit IRC22:52
*** zaitcev has joined #openstack-swift22:53
*** fleducquede_ has joined #openstack-swift22:53
*** haomaiwa_ has joined #openstack-swift22:53
*** pconstantine has joined #openstack-swift22:53
*** otoolee has joined #openstack-swift22:53
*** rpedde has joined #openstack-swift22:53
*** cschwede has joined #openstack-swift22:53
*** swat30 has joined #openstack-swift22:53
*** acoles_away has joined #openstack-swift22:53
*** sandywalsh has joined #openstack-swift22:53
*** sets mode: +vv zaitcev acoles_away22:53
*** kevinc_ has joined #openstack-swift22:53
*** clayg has joined #openstack-swift22:53
*** ekarlso has joined #openstack-swift22:53
*** dfg has joined #openstack-swift22:53
*** jokke_ has joined #openstack-swift22:53
*** redbo has joined #openstack-swift22:53
*** thurloat has joined #openstack-swift22:53
*** sets mode: +vvv clayg dfg redbo22:53
*** Mikalv has quit IRC22:54
*** Mikalv has joined #openstack-swift22:54
*** patchbot has quit IRC22:54
*** zacksh has quit IRC22:54
*** ryao has quit IRC22:54
*** mtreinish has quit IRC22:54
*** zigo has quit IRC22:54
*** MooingLemur has quit IRC22:54
*** openfly has quit IRC22:54
*** bobby_ has quit IRC22:54
*** bgmccollum has quit IRC22:54
*** pandemicsyn has quit IRC22:54
*** nottrobin has quit IRC22:54
*** StevenK has quit IRC22:54
*** dmorita has quit IRC22:55
*** sudorandom has quit IRC22:55
*** AbyssOne has quit IRC22:55
*** tanee2 has quit IRC22:55
*** ondergetekende_ has quit IRC22:55
*** infotection has quit IRC22:55
*** joearnold has quit IRC22:55
*** ciypro|afk has quit IRC22:55
*** patchbot has joined #openstack-swift22:55
*** zacksh has joined #openstack-swift22:55
*** ryao has joined #openstack-swift22:55
*** mtreinish has joined #openstack-swift22:55
*** zigo has joined #openstack-swift22:55
*** MooingLemur has joined #openstack-swift22:55
*** bobby_ has joined #openstack-swift22:55
*** openfly has joined #openstack-swift22:55
*** bgmccollum has joined #openstack-swift22:55
*** pandemicsyn has joined #openstack-swift22:55
*** nottrobin has joined #openstack-swift22:55
*** StevenK has joined #openstack-swift22:55
*** ryao has quit IRC22:55
*** ryao has joined #openstack-swift22:55
MooingLemurtorgomatic: thanks for the answer, I guess I wanted to bend write_affinity beyond its intended purpose.  Looks problematic. :)22:56
*** dmorita has joined #openstack-swift22:56
*** sudorandom has joined #openstack-swift22:56
*** AbyssOne has joined #openstack-swift22:56
*** tanee2 has joined #openstack-swift22:56
*** ondergetekende_ has joined #openstack-swift22:56
*** infotection has joined #openstack-swift22:56
*** joearnold has joined #openstack-swift22:56
*** ciypro|afk has joined #openstack-swift22:56
torgomaticMooingLemur: Yep :) I think you really want an HTTP write-back caching proxy that handles PUTs; combine that with tempurls for the PUT requests, and it might stand a chance of workign22:57
torgomaticof course, I don't know if anyone's written such a thing; all the HTTP caches I know are focused on GET22:57
*** Tyger_ has joined #openstack-swift22:58
*** Tyger has quit IRC23:00
* MooingLemur ponders23:05
*** elambert has joined #openstack-swift23:05
MooingLemuryeah, if I could have a generic writeback HTTP cache, that speaks swift auth. :)23:06
MooingLemur(so that it can give the client a reasonable return code based on whether it'd be likely to succeed on swift itself)23:08
*** zul has quit IRC23:17
torgomaticMooingLemur: well, that's why you'd use tempurl for the PUTs; it's a different auth mechanism that's more likely to work :)23:23
MooingLemurtorgomatic: Well, this idea came up to placate users complaining about ingest speeds.  The easiest remedy is to have them parallelize uploads more, because they are geographically kinda far from the proxy, and also because the object sizes are small.  I was just trying to come up with a solution that would improve the case where the user made no changes to their process.23:29
torgomaticMooingLemur: ah... yeah, either move the cluster closer to the users, or have them do more stuff in parallel23:30
torgomaticthere's also the bulk uploads; I think those can take .tar.gz so you get some compression going, if the users are bandwidth-constrained23:31
*** zul has joined #openstack-swift23:31
*** kevinc_ has quit IRC23:35
*** occupant has quit IRC23:48
*** occupant has joined #openstack-swift23:49

Generated by 2.14.0 by Marius Gedminas - find it at!