| *** tongli has joined #openstack-swift | 00:02 | |
| *** fbo_away is now known as fbo | 00:02 | |
| *** tongli has quit IRC | 00:03 | |
| notmyname | can someone else +2 https://review.openstack.org/#/c/78070/ (it's a really simple change) | 00:03 |
|---|---|---|
| notmyname | well, review it and add your +2 if you like it :-) | 00:03 |
| *** Fin1te has joined #openstack-swift | 00:04 | |
| *** shri has quit IRC | 00:09 | |
| *** miurahr has quit IRC | 00:10 | |
| *** sungju has quit IRC | 00:13 | |
| *** shri has joined #openstack-swift | 00:14 | |
| *** fbo is now known as fbo_away | 00:17 | |
| *** haomaiwa_ has joined #openstack-swift | 00:21 | |
| *** mlipchuk has quit IRC | 00:24 | |
| *** miurahr has joined #openstack-swift | 00:25 | |
| *** miurahr has left #openstack-swift | 00:26 | |
| *** Fin1te has quit IRC | 00:27 | |
| torgomatic | anyone know why normalize_timestamp uses a 10.5 format and not 10.6? | 00:34 |
| torgomatic | I thought we had microsecond resolution | 00:35 |
| torgomatic | I looked in the git history, but it all goes back to the initial commit | 00:37 |
| *** Midnightmyth has quit IRC | 00:40 | |
| *** Fin1te has joined #openstack-swift | 00:40 | |
| *** openstackgerrit has joined #openstack-swift | 00:40 | |
| openstackgerrit | A change was merged to openstack/swift: Send policy index on async update https://review.openstack.org/71702 | 00:42 |
| openstackgerrit | A change was merged to openstack/swift: Send policy index to container on sync update https://review.openstack.org/71703 | 00:42 |
| *** miurahr has joined #openstack-swift | 00:45 | |
| *** Fin1te has quit IRC | 00:48 | |
| *** miurahr has quit IRC | 00:50 | |
| *** keving1 has quit IRC | 00:51 | |
| openstackgerrit | A change was merged to openstack/swift: Fix a couple typos https://review.openstack.org/77041 | 00:53 |
| openstackgerrit | A change was merged to openstack/swift: Make internal client work with conf.d https://review.openstack.org/78133 | 00:53 |
| openstackgerrit | A change was merged to openstack/python-swiftclient: Make the help strings constant https://review.openstack.org/74578 | 00:54 |
| *** miurahr has joined #openstack-swift | 00:55 | |
| *** miurahr has left #openstack-swift | 00:55 | |
| openstackgerrit | A change was merged to openstack/python-swiftclient: Help string format persistent https://review.openstack.org/76630 | 00:57 |
| *** Midnightmyth has joined #openstack-swift | 00:57 | |
| *** Dharmit has joined #openstack-swift | 01:06 | |
| *** yuan has joined #openstack-swift | 01:12 | |
| *** kun_huang has joined #openstack-swift | 01:16 | |
| *** peluse has quit IRC | 01:16 | |
| *** peluse has joined #openstack-swift | 01:17 | |
| *** chuck__ has joined #openstack-swift | 01:21 | |
| *** haomaiwa_ has quit IRC | 01:32 | |
| *** haomaiwang has joined #openstack-swift | 01:33 | |
| *** nosnos has joined #openstack-swift | 01:35 | |
| *** haomaiwang has quit IRC | 01:37 | |
| *** shri has quit IRC | 01:40 | |
| *** haomaiwa_ has joined #openstack-swift | 01:43 | |
| *** haomaiwa_ has quit IRC | 01:43 | |
| *** haomaiwang has joined #openstack-swift | 01:46 | |
| *** yuan has quit IRC | 01:52 | |
| openstackgerrit | A change was merged to openstack/swift: sanity check copy tests https://review.openstack.org/77811 | 02:06 |
| *** yuan has joined #openstack-swift | 02:13 | |
| *** csd has quit IRC | 02:17 | |
| *** chandan_kumar has joined #openstack-swift | 02:21 | |
| *** bada_ has joined #openstack-swift | 02:28 | |
| *** bada has quit IRC | 02:28 | |
| *** Midnightmyth has quit IRC | 02:30 | |
| *** keving1 has joined #openstack-swift | 02:40 | |
| *** bada has joined #openstack-swift | 02:50 | |
| *** keving1 has quit IRC | 02:50 | |
| *** bada_ has quit IRC | 02:52 | |
| portante | torgomatic: I had wondered that same thing a while back, as it means there is a slightly higher possibility of false collisions | 02:56 |
| *** chandan_kumar has quit IRC | 02:59 | |
| *** madhuri has quit IRC | 03:01 | |
| *** Slidey has quit IRC | 03:07 | |
| *** kun_huang has quit IRC | 03:16 | |
| *** pberis has quit IRC | 03:25 | |
| *** pberis has joined #openstack-swift | 03:28 | |
| *** haomaiw__ has joined #openstack-swift | 03:36 | |
| *** haomaiwang has quit IRC | 03:39 | |
| *** keving1 has joined #openstack-swift | 03:46 | |
| *** keving1 has quit IRC | 03:55 | |
| *** fifieldt has joined #openstack-swift | 03:55 | |
| *** bada has quit IRC | 04:00 | |
| *** vu_ has quit IRC | 04:07 | |
| *** vu has quit IRC | 04:07 | |
| *** csd has joined #openstack-swift | 04:08 | |
| *** csd has quit IRC | 04:12 | |
| *** haomaiw__ has quit IRC | 04:13 | |
| *** haomaiwang has joined #openstack-swift | 04:14 | |
| *** haomaiwang has quit IRC | 04:15 | |
| *** haomaiwa_ has joined #openstack-swift | 04:15 | |
| *** zaitcev has quit IRC | 04:30 | |
| *** ppai has joined #openstack-swift | 04:39 | |
| *** bvandenh has joined #openstack-swift | 04:41 | |
| *** madhuri has joined #openstack-swift | 05:07 | |
| madhuri | Hi, can anyone tell me what happens when metadata for object is set to empty in POST method? | 05:09 |
| *** kun_huang has joined #openstack-swift | 05:13 | |
| *** keving1 has joined #openstack-swift | 05:13 | |
| portante | do you mean, the x-object-metadata is present but empty? | 05:18 |
| portante | madhuri, or do you mean when the header is not included at all in the REST API request? | 05:18 |
| *** Slidey has joined #openstack-swift | 05:20 | |
| madhuri | portante: No, I include it in header and pass empty value for it. | 05:21 |
| madhuri | portante: When I set it to empty and HEAD it, it is not displayed in output. But resp.headers contain it with empty value. Is my understanding correct? | 05:21 |
| portante | I'll have to check | 05:22 |
| madhuri | portante: Ok | 05:22 |
| madhuri | portante: http://pastebin.com/52DnWZ4z | 05:24 |
| openstackgerrit | Peter Portante proposed a change to openstack/swift: Add missing swift constraint for functional tests https://review.openstack.org/78522 | 05:26 |
| portante | torgomatic, notmyname, clayg: can you guys consider the above? | 05:27 |
| portante | I'd like to see that fix land before we try to work through the in-process functional tests | 05:27 |
| portante | madhuri: okay | 05:27 |
| portante | so a POST of an object replaces all the metadata entirely. | 05:27 |
| portante | so it is effectively deleting the metadata when you include the header with an empty value | 05:28 |
| portante | but I don't see the empty header in the example you gave above | 05:29 |
| madhuri | portante: Yes | 05:30 |
| madhuri | portante: but the response header include the metadata with a empty value | 05:30 |
| madhuri | portante: I am writing a test for it. And the response header is : Response header is {'Content-Length': '0', 'Last-Modified': 'Thu, 06 Mar 2014 05:15:32 GMT', 'Etag': '"d41d8cd98f00b204e9800998ecf8427e"', 'X-Timestamp': '1394082931.50516', 'X-Object-Meta-1': '', 'Content-Type': 'application/x-test'} | 05:31 |
| portante | madhuri: I was confused, because I don't see that header in your pastebin | 05:35 |
| portante | ah | 05:35 |
| portante | you mean the response header of the POST operation? | 05:35 |
| madhuri | portante: No, response header in HEAD method. | 05:36 |
| portante | okay, so I am back to being confused | 05:36 |
| *** csd has joined #openstack-swift | 05:37 | |
| portante | the pastebin you posted above seems to show no X-Object-Meta-test header in the HEAD response | 05:37 |
| madhuri | portante: The POST header is just this {'Content-Length': '76', 'Content-Type': 'text/html; charset=UTF-8'} | 05:37 |
| portante | okay | 05:37 |
| madhuri | portante: Yes. My understanding is same too and according to me it is working properly. | 05:38 |
| madhuri | portante: But the response header includes it. | 05:38 |
| portante | I am still confused as to what is happening | 05:38 |
| portante | one example you are showing me seems to NOT have it, the other example does have it | 05:39 |
| portante | do you agree? | 05:39 |
| madhuri | portante: Yes | 05:39 |
| madhuri | portante: I have written a test for it i.e setting metadata to empty | 05:40 |
| madhuri | portante: acoording to me the metadata must be deleted and must not show up in HEAD method. Right?? | 05:40 |
| portante | It looks like the code is going to store the blank value, and it should be returned that way | 05:42 |
| *** keving1 has quit IRC | 05:48 | |
| madhuri | portante: Yes emtpy value is stored. But it doesn't show up in output of HEAD. So why is it stored? | 05:58 |
| *** gyee has quit IRC | 06:00 | |
| *** chandan_kumar has joined #openstack-swift | 06:04 | |
| portante | it is stored because it is present in the headers, and so we store it as is | 06:09 |
| portante | why it does not show up in the HEAD response is another question | 06:09 |
| madhuri | portante: Ok. | 06:10 |
| madhuri | portante: But this doen't follow in account metadata. | 06:10 |
| madhuri | portante: If we set account metadata to empty, it is not present in response header. | 06:11 |
| madhuri | portante: :( | 06:11 |
| portante | ah | 06:11 |
| portante | okay, so container and account metadata is handled differently than object | 06:11 |
| portante | for account and container metadata, a POST only updates the metadata referenced, where as a POST on an object must include all the metadata in the request | 06:12 |
| portante | madhuri: | 06:12 |
| madhuri | portante: Thanks a tonne. But it is confusing a bit | 06:13 |
| madhuri | portante: One more question | 06:16 |
| madhuri | portante: https://github.com/openstack/swift/blob/master/swift/account/server.py#L100 | 06:16 |
| madhuri | portante: https://github.com/openstack/swift/blob/master/swift/account/server.py#L80 | 06:17 |
| madhuri | portante: Why we are checking the status of account twice? | 06:17 |
| portante | madhuri: not sure why exactly, would have to look into it | 06:23 |
| *** nshaikh has joined #openstack-swift | 06:23 | |
| madhuri | portante: So can I add a test for empty metadata for object? | 06:25 |
| portante | sure, go ahead a propose a patch so that we can discuss the details concretely | 06:28 |
| portante | I'm gonna sign off now, g'nite | 06:28 |
| madhuri | portante: Thank you :) | 06:29 |
| madhuri | portante: Good night. | 06:29 |
| *** csd has quit IRC | 06:35 | |
| *** haomaiw__ has joined #openstack-swift | 06:35 | |
| *** haomaiwa_ has quit IRC | 06:39 | |
| openstackgerrit | Peter Portante proposed a change to openstack/swift: In-process swift server for functional tests https://review.openstack.org/66108 | 06:55 |
| *** otoolee has quit IRC | 07:02 | |
| *** haomaiw__ has quit IRC | 07:04 | |
| *** otoolee has joined #openstack-swift | 07:04 | |
| *** haomaiwa_ has joined #openstack-swift | 07:04 | |
| openstackgerrit | Madhuri Kumari proposed a change to openstack/swift: Added a test for empty metadata https://review.openstack.org/78530 | 07:07 |
| *** kun_huang has quit IRC | 07:13 | |
| *** saju_m has joined #openstack-swift | 07:22 | |
| *** basha has joined #openstack-swift | 07:57 | |
| *** saju_m has quit IRC | 08:06 | |
| *** saju_m has joined #openstack-swift | 08:18 | |
| *** saju_m has quit IRC | 08:24 | |
| *** openstackgerrit has quit IRC | 08:34 | |
| *** openstackgerrit has joined #openstack-swift | 08:34 | |
| *** saju_m has joined #openstack-swift | 08:36 | |
| *** basha has quit IRC | 08:39 | |
| openstackgerrit | Christian Schwede proposed a change to openstack/swift: Add containeralias middleware https://review.openstack.org/62494 | 08:46 |
| *** basha has joined #openstack-swift | 08:47 | |
| *** basha has quit IRC | 08:52 | |
| *** basha has joined #openstack-swift | 08:55 | |
| *** nacim has joined #openstack-swift | 09:05 | |
| *** tanee-away is now known as tanee | 09:08 | |
| *** tanee is now known as tanee-away | 09:09 | |
| *** mlipchuk has joined #openstack-swift | 09:11 | |
| *** saju_m has quit IRC | 09:12 | |
| *** tanee-away is now known as tanee | 09:13 | |
| *** bada has joined #openstack-swift | 09:14 | |
| *** psharma has joined #openstack-swift | 09:16 | |
| *** fbo_away is now known as fbo | 09:17 | |
| *** mkerrin has quit IRC | 09:18 | |
| *** Dharmit has quit IRC | 09:19 | |
| *** mkerrin has joined #openstack-swift | 09:19 | |
| *** mmcardle has joined #openstack-swift | 09:30 | |
| *** Anju has left #openstack-swift | 09:33 | |
| *** kun_huang has joined #openstack-swift | 09:40 | |
| *** basha has quit IRC | 09:42 | |
| *** foexle has joined #openstack-swift | 09:53 | |
| *** jamieh has joined #openstack-swift | 09:57 | |
| *** saju_m has joined #openstack-swift | 09:58 | |
| *** Midnightmyth has joined #openstack-swift | 10:00 | |
| openstackgerrit | Madhuri Kumari proposed a change to openstack/swift: Added a test for empty metadata https://review.openstack.org/78530 | 10:03 |
| *** ccorrigan has quit IRC | 10:03 | |
| *** mlipchuk has left #openstack-swift | 10:04 | |
| *** nshaikh has quit IRC | 10:05 | |
| *** mlipchuk has joined #openstack-swift | 10:13 | |
| *** jamieh has quit IRC | 10:14 | |
| *** jamieh has joined #openstack-swift | 10:15 | |
| *** bvandenh has quit IRC | 10:36 | |
| *** saju_m has quit IRC | 10:46 | |
| *** bvandenh has joined #openstack-swift | 11:03 | |
| *** saju_m has joined #openstack-swift | 11:03 | |
| *** haomaiw__ has joined #openstack-swift | 11:07 | |
| *** haomaiwa_ has quit IRC | 11:07 | |
| *** tanee is now known as tanee-away | 11:38 | |
| *** kun_huang has quit IRC | 11:40 | |
| *** ppai has quit IRC | 12:01 | |
| *** lpabon has joined #openstack-swift | 12:03 | |
| *** mmcardle has quit IRC | 12:03 | |
| *** miurahr has joined #openstack-swift | 12:08 | |
| *** miurahr has quit IRC | 12:12 | |
| *** ppai has joined #openstack-swift | 12:12 | |
| *** bvandenh has quit IRC | 12:18 | |
| *** mkollaro has joined #openstack-swift | 12:19 | |
| *** bada has quit IRC | 12:21 | |
| *** j_king has quit IRC | 12:34 | |
| *** kun_huang has joined #openstack-swift | 12:46 | |
| *** foexle has quit IRC | 12:55 | |
| *** judd7 has quit IRC | 12:57 | |
| *** tanee-away is now known as tanee | 12:59 | |
| *** ppai has quit IRC | 13:03 | |
| *** PradeepChandani has joined #openstack-swift | 13:04 | |
| *** psharma has quit IRC | 13:16 | |
| *** saju_m has quit IRC | 13:27 | |
| *** chuck__ has quit IRC | 13:33 | |
| *** saurabh_ has joined #openstack-swift | 13:43 | |
| *** saurabh_ has joined #openstack-swift | 13:43 | |
| *** lpabon has quit IRC | 13:47 | |
| *** changbl has quit IRC | 14:14 | |
| *** judd7 has joined #openstack-swift | 14:20 | |
| *** vu_ has joined #openstack-swift | 14:22 | |
| *** j_king has joined #openstack-swift | 14:23 | |
| *** kun_huang has quit IRC | 14:27 | |
| *** hurricanerix has joined #openstack-swift | 14:32 | |
| *** fifieldt has quit IRC | 14:35 | |
| *** mkollaro has quit IRC | 14:41 | |
| *** zul has quit IRC | 14:49 | |
| *** kun_huang has joined #openstack-swift | 14:50 | |
| *** zul has joined #openstack-swift | 14:52 | |
| *** nosnos has quit IRC | 14:55 | |
| *** SkyRocknRoll__ has joined #openstack-swift | 14:56 | |
| *** occupant has quit IRC | 14:57 | |
| *** occupant has joined #openstack-swift | 14:57 | |
| *** tdasilva has joined #openstack-swift | 14:58 | |
| *** mkollaro has joined #openstack-swift | 15:00 | |
| peluse | portante: you there? | 15:06 |
| portante | yes | 15:07 |
| portante | peluse: good morning | 15:07 |
| peluse | good morning! | 15:08 |
| peluse | hey, for that split exception we were talking about... I'm just catching it in the rcvr now and both logging a warn and raising, think that's good? (pasting example) | 15:09 |
| portante | k | 15:09 |
| portante | hmm | 15:09 |
| peluse | Mar 6 08:13:12 sw2 object-server: ERROR: Replication parameter mismatch, possibly using policies with partially upgraded cluster? | 15:09 |
| peluse | Mar 6 08:13:12 sw2 object-server: 127.0.0.1/sdb2/684 EXCEPTION in replication.Receiver: #012Traceback (most recent call last):#012 File "/home/peluse/swift/swift/obj/ssync_receiver.py", line 104, in __call__#012 for data in self.missing_check():#012 File "/home/peluse/swift/swift/obj/ssync_receiver.py", line 231, in missing_check#012 [urllib.unquote(v) for v in line.split()] | 15:09 |
| peluse | #012ValueError: need more than 2 values to unpack | 15:09 |
| portante | so this is a new receiver catching a bad message from an old sender, right? | 15:10 |
| peluse | this is an old rcvd getting a msg from a new sender | 15:10 |
| *** saurabh_ has left #openstack-swift | 15:10 | |
| peluse | rcvr | 15:10 |
| portante | but can we really change the old rcvr code? | 15:11 |
| portante | are you making a change on master for the rcvr? | 15:11 |
| peluse | crap, have to take kids to school... back in a bit (sorry) | 15:12 |
| portante | k, no problem | 15:12 |
| peluse | back for a few... so that change would cover new rcvr old sender (I wrote it backwards) and a similar change can be added to the sender to catch new sender old rcvr. We can cover it both directions | 15:22 |
| portante | yes | 15:22 |
| peluse | OK, cool :) | 15:23 |
| portante | except you don't need to do a warning that way, I don't think | 15:23 |
| portante | meaning | 15:23 |
| portante | we can support old -> new, but it is the new -> old case that is the one we want a warning for | 15:23 |
| portante | does that make sense? | 15:23 |
| peluse | it does.. thanks | 15:24 |
| *** tanee is now known as tanee-away | 15:24 | |
| peluse | so for the rcvr, I should just catch the error and set the index to 0 (not raise or even wwarn) then right? | 15:25 |
| portante | yes | 15:25 |
| peluse | got it, thanks. how can something so simple be so confusing :) | 15:25 |
| *** ccorrigan has joined #openstack-swift | 15:25 | |
| portante | for a new sender, if the sent message is returned with a 500, then we might want to warn about a mismatch | 15:25 |
| portante | ;) | 15:26 |
| peluse | cool, look for the updated patch a bit later. Have a few hrs of mtgs first then I'll wrap it up | 15:26 |
| *** tanee-away is now known as tanee | 15:27 | |
| portante | k, thx | 15:27 |
| *** piyush has joined #openstack-swift | 15:27 | |
| peluse | BTW, I'm going to setup a cluster and see if I can't perform a rolling upgrade going from master to featre/ec and see what else crops up, I have a feeling there are a few situations like this that haven't been uncovered yet | 15:31 |
| portante | great idea | 15:31 |
| *** tanee is now known as tanee-away | 15:32 | |
| *** tanee-away is now known as tanee | 15:35 | |
| *** tanee is now known as tanee-away | 15:35 | |
| luisbg | morning | 15:39 |
| portante | good morning | 15:39 |
| peluse | morning | 15:40 |
| luisbg | portante, and peluse, :) | 15:41 |
| portante | luisbg: good morning | 15:44 |
| *** judd7 has quit IRC | 15:47 | |
| *** judd7 has joined #openstack-swift | 15:48 | |
| *** judd7 has quit IRC | 15:50 | |
| *** judd7 has joined #openstack-swift | 15:51 | |
| *** hurrican_ has joined #openstack-swift | 15:55 | |
| *** hurrican_ has quit IRC | 15:57 | |
| *** hurricanerix has quit IRC | 15:58 | |
| *** hurricanerix has joined #openstack-swift | 15:58 | |
| *** tanee-away is now known as tanee | 15:59 | |
| *** tanee is now known as tanee-away | 15:59 | |
| *** hurricanerix has quit IRC | 16:02 | |
| peluse | portante: on the sender side (new sender old rcvr) I don't get any specific error info back when an exception happens on the rcvr, Looks like if anything goes wrong on the rcv side the sender simply does exceptions.ReplicationException('Early disconnect') | 16:03 |
| *** kun_huang has quit IRC | 16:07 | |
| portante | peluse: what does that translate to for an HTTP response status code? | 16:07 |
| *** tanee-away is now known as tanee | 16:08 | |
| peluse | checking... | 16:11 |
| peluse | there's nothing in the response body | 16:12 |
| portante | so, what about the HTTP response headers? | 16:13 |
| portante | and what is the response status code? | 16:13 |
| *** chandan_kumar has quit IRC | 16:13 | |
| *** kun_huang has joined #openstack-swift | 16:15 | |
| openstackgerrit | Madhuri Kumari proposed a change to openstack/swift: Added a test for empty metadata https://review.openstack.org/78530 | 16:16 |
| peluse | portante: can't seem to get that info | 16:20 |
| *** changbl has joined #openstack-swift | 16:20 | |
| portante | hmm | 16:23 |
| peluse | see if this makes any sense to you - this is new sender to old rcvr error: https://gist.github.com/peluse/9393383 | 16:24 |
| *** lpabon has joined #openstack-swift | 16:25 | |
| portante | peluse, we have to find out about that exception and what that leads to | 16:26 |
| portante | can you turn on request logging for the object server? | 16:26 |
| portante | it appears to be turned off | 16:27 |
| *** hurricanerix has joined #openstack-swift | 16:28 | |
| peluse | I thought it was on by default, just set it explciitly but don't see any difference. The exception in what I posted is expected though as I'm creating the scenario new-->old, the question is how to detect it from the sender and all I get is a blank response and a rsp code of 200 | 16:33 |
| *** vu_ has quit IRC | 16:35 | |
| *** mlipchuk has quit IRC | 16:39 | |
| portante | hmmm | 16:40 |
| *** basha has joined #openstack-swift | 16:40 | |
| portante | peluse: I am tracking down another work item, can I look at this later if nobody else gets to it? | 16:40 |
| *** mkollaro has quit IRC | 16:42 | |
| peluse | portante: no problem | 16:44 |
| *** Dharmit has joined #openstack-swift | 16:44 | |
| peluse | I'll keep digging too | 16:44 |
| *** mkollaro has joined #openstack-swift | 16:44 | |
| *** lpabon has quit IRC | 16:45 | |
| *** basha has quit IRC | 16:45 | |
| peluse | I'll also check with gholt later on as well... | 16:45 |
| portante | k | 16:46 |
| *** zaitcev has joined #openstack-swift | 16:49 | |
| *** ChanServ sets mode: +v zaitcev | 16:49 | |
| *** mkollaro has quit IRC | 16:53 | |
| *** bada has joined #openstack-swift | 16:56 | |
| *** Dharmit has quit IRC | 17:00 | |
| *** basha has joined #openstack-swift | 17:03 | |
| *** bada has quit IRC | 17:05 | |
| *** mkollaro has joined #openstack-swift | 17:09 | |
| *** gyee has joined #openstack-swift | 17:10 | |
| *** basha has quit IRC | 17:22 | |
| *** SkyRocknRoll__ has quit IRC | 17:27 | |
| notmyname | hello world | 17:28 |
| ccorrigan | :notmyname :) | 17:29 |
| *** jamieh has quit IRC | 17:30 | |
| *** jamieh_ has joined #openstack-swift | 17:35 | |
| *** vu_ has joined #openstack-swift | 17:42 | |
| *** vu_ has quit IRC | 17:46 | |
| *** vu_ has joined #openstack-swift | 17:46 | |
| *** csd has joined #openstack-swift | 17:50 | |
| portante | notmyname: hello worldling | 17:55 |
| *** lpabon has joined #openstack-swift | 17:59 | |
| *** shri has joined #openstack-swift | 18:01 | |
| *** taras__ has joined #openstack-swift | 18:02 | |
| taras__ | hi guys, i'm a firefox dev who does a lot with sqlite. i read that swift uses sqlite as a key/value mapping to match keys and objects | 18:02 |
| taras__ | sounds like there might be some similar problems to what i fixed in ff in swift | 18:03 |
| notmyname | taras__: sortof. we use sqlite for keeping listings of containers and objects | 18:03 |
| taras__ | could someone do me a favor and run filefrag on the sqlite db for a production bucket with a large number of objects? | 18:03 |
| taras__ | s/production/anything that's been in use for a while/ | 18:04 |
| taras__ | when i last tested this, xfs and filefrag had some serious fragmentation issues that would kill sqlite throughput | 18:05 |
| taras__ | xfs and sqlite i mean | 18:05 |
| *** chandan_kumar has joined #openstack-swift | 18:07 | |
| taras__ | notmyname: listings of objects is the case that might be hurting | 18:07 |
| taras__ | this is based on what i read in http://engineering.spilgames.com/openstack-swift-lots-small-files/ | 18:08 |
| openstackgerrit | Peter Portante proposed a change to openstack/swift: Add missing swift constraint for functional tests https://review.openstack.org/78522 | 18:10 |
| *** tanee is now known as tanee-away | 18:11 | |
| openstackgerrit | Peter Portante proposed a change to openstack/swift: In-process swift server for functional tests https://review.openstack.org/66108 | 18:12 |
| *** kun_huang has quit IRC | 18:15 | |
| *** piyush has quit IRC | 18:16 | |
| *** nacim has quit IRC | 18:19 | |
| *** dmsimard has joined #openstack-swift | 18:19 | |
| *** jogo has joined #openstack-swift | 18:21 | |
| jogo | so we are seeing a new swift failure in gate http://paste.openstack.org/show/72806/ | 18:23 |
| jogo | wondering if this has been seen before and if there is a bug | 18:23 |
| notmyname | jogo: not that I know of, but I'm looking | 18:25 |
| jogo | notmyname: cool, also not sure if this is a tempest issue or swift | 18:25 |
| notmyname | jogo: not a lot of info there | 18:26 |
| jogo | notmyname: no :/ | 18:26 |
| *** mandarine has quit IRC | 18:27 | |
| jogo | notmyname: but there is a recent spike in hits | 18:27 |
| clayg | jogo! | 18:27 |
| jogo | clayg! | 18:28 |
| * jogo waves at clayg | 18:28 | |
| notmyname | jogo: I can't really tell too much from that output. looks like it's testing something to do with quotas. but I don't know what test is failing. do you have a link to the actual test that is failing? | 18:28 |
| jogo | http://logs.openstack.org/79/71479/16/check/check-tempest-dsvm-postgres-full/309394e/logs/testr_results.html.gz | 18:29 |
| jogo | notmyname: not sure if that helps | 18:29 |
| notmyname | jogo: that's exactly the same thing | 18:29 |
| notmyname | where's the tempest code? /me goes to look | 18:29 |
| *** gyee has quit IRC | 18:30 | |
| jogo | notmyname: only one change to tempest/api/object_storage/test_account_quotas_negative.py recently and it doesn't look related | 18:30 |
| jogo | it landed on the 1st | 18:31 |
| portante | are all the servers up and running? | 18:31 |
| notmyname | hmm..my local copy is 1523 commits out of date... | 18:31 |
| portante | jogo, notmyname, we have seen times in the past where the server processes fail to start during some tests | 18:32 |
| notmyname | ya | 18:32 |
| notmyname | also account quotas rely on the updater process (so aren't immediately updated) | 18:32 |
| *** joeljwright has joined #openstack-swift | 18:32 | |
| portante | notmyname: the updater process meaning the container updater | 18:34 |
| portante | right? | 18:35 |
| portante | not the object updater, or do you mean both | 18:35 |
| jogo | notmyname portante: filed a swift/tempest bug for this https://bugs.launchpad.net/tempest/+bug/1288918 | 18:35 |
| notmyname | yes, eventually. so that the used bytes in the account db are right and a quota can be enforced (assuming the object isn't bigger than the quota) | 18:35 |
| notmyname | hmm...looks like tempest is setting the quota to 20 bytes and then trying to upload a 30 byte object | 18:37 |
| jogo | notmyname: are you fine with using this as a fingerprint until we know more: message:" File \"tempest/api/object_storage/test_account_quotas_negative.py\"" AND message:" in test_upload_large_object" AND filename:"console.html" | 18:39 |
| jogo | for elastic-recheck | 18:40 |
| notmyname | jogo: I have no opinions on that | 18:43 |
| jogo | kk | 18:44 |
| *** chuck__ has joined #openstack-swift | 18:45 | |
| *** mkollaro has quit IRC | 18:47 | |
| joeljwright | hi everyone, I wanted to get some advice about the blueprint process for python-swiftclient | 18:47 |
| joeljwright | I've got a few projects that have been accessing swift using python, but there was a lot of logic buried in bin/swift that wasn't importable | 18:48 |
| joeljwright | I refactored that code out into a swift service and registered a blueprint | 18:49 |
| clayg | good god why didn't we make 'swift capabilities' just 'swift info' - heck it even matches the /info url better? | 18:49 |
| joeljwright | thought it might be useful to others | 18:49 |
| notmyname | practice spelling capibili...capabilety...capabilities | 18:50 |
| portante | capibara? | 18:50 |
| notmyname | joeljwright: sounds good. love to see the patches. don't worry about the blueprints. among other things, LP has no alerting for when blueprints are registered, so I doubt anyone has even looked at it | 18:50 |
| joeljwright | ah ok | 18:51 |
| StevenK | Hm, there is blueprint notification. I thought notify on create worked. Oh well. | 18:51 |
| joeljwright | it's unfinished, and needs more work before it's ready, but I'm willing to put in the time if it's interesting | 18:52 |
| openstackgerrit | Clay Gerrard proposed a change to openstack/python-swiftclient: add "info" as an alias to "capabilities" https://review.openstack.org/78710 | 18:52 |
| clayg | ^ so much better | 18:52 |
| joeljwright | was hoping to guage interest before I spend a lot of time polishing and writing tests | 18:52 |
| joeljwright | notmyname: have a forked github repo here: https://github.com/sohonetlabs/python-swiftclient | 18:53 |
| notmyname | joeljwright: https://github.com/sohonetlabs/python-swiftclient/commit/357cd3cb7d22cf020119d250fded4b7060a45dfa ? | 18:54 |
| notmyname | StevenK: I'd be happy to be corrected, if you can point me to the setting to get alerts on new blueprints | 18:55 |
| StevenK | notmyname: I was thinking about the code itself. Notifications tend to have no settings at all. | 18:56 |
| joeljwright | notmyname: unfortunately it's evolved over a long period and gone through 2 versions | 18:57 |
| joeljwright | vast majority of the logic in bin/swift has moved into swiftclient/service.py | 18:57 |
| joeljwright | so bin/swift is just arg parsing and output | 18:57 |
| *** chuck__ has quit IRC | 18:57 | |
| clayg | joeljwright: that all looks pretty good to me | 18:58 |
| clayg | just put that diff up as a patch | 18:58 |
| notmyname | joeljwright: in general, I think we'd all support that. among other things, it would really help out for test coverage. there has been some similar work in the swift repo too (but using a "cli" directory rather than a services module--see https://github.com/openstack/swift/tree/master/swift/cli) | 18:58 |
| notmyname | jogo: FWIW, it works for me in my SAIO. I set the quota to 20 and upload ~40 bytes and get a 413 (as expected) | 18:59 |
| jogo | notmyname: not surprised, this works most of the time in gate too | 19:01 |
| jogo | notmyname: only 4 failures or so in last 24 hours | 19:01 |
| jogo | so very infrequent | 19:01 |
| notmyname | also, FWIW, I had no background jobs running (/cc portante) | 19:01 |
| joeljwright | clayg: the problem with a diff atm is that I forked the python-swiftclient code in Nov last year, so there have obviously been a lot of changes since then | 19:05 |
| notmyname | yup | 19:05 |
| joeljwright | notmyname: it's good to know that the general idea is one that would be supported | 19:05 |
| joeljwright | I'll certainly spend some time rebasing onto a newer version, writing tests and cleaning up | 19:06 |
| joeljwright | unfortunately I'm away travelling for the next 2 weeks, but I'll continue working on it when I get back | 19:07 |
| *** mandarine has joined #openstack-swift | 19:16 | |
| *** piyush has joined #openstack-swift | 19:17 | |
| *** piyush1 has joined #openstack-swift | 19:19 | |
| *** piyush has quit IRC | 19:21 | |
| openstackgerrit | Fabien Boucher proposed a change to openstack/python-swiftclient: Allow get_account and get_container to return an iterator https://review.openstack.org/74845 | 19:24 |
| *** bada has joined #openstack-swift | 19:27 | |
| *** gyee has joined #openstack-swift | 19:29 | |
| *** kun_huang has joined #openstack-swift | 19:39 | |
| *** mkollaro has joined #openstack-swift | 19:46 | |
| *** bada has quit IRC | 19:48 | |
| openstackgerrit | paul luse proposed a change to openstack/swift: Add Storage Policy Support to ssync https://review.openstack.org/65347 | 19:51 |
| *** judd7_ has joined #openstack-swift | 19:51 | |
| peluse | gholt: you out there? | 19:51 |
| *** judd7 has quit IRC | 19:51 | |
| *** openstackgerrit has quit IRC | 20:00 | |
| *** openstackgerrit_ has joined #openstack-swift | 20:00 | |
| *** openstackgerrit_ is now known as openstackgerrit | 20:01 | |
| *** joeljwright has left #openstack-swift | 20:01 | |
| *** judd7_ has quit IRC | 20:02 | |
| *** erlon_ has joined #openstack-swift | 20:03 | |
| *** vu_ has quit IRC | 20:03 | |
| *** erlon has quit IRC | 20:03 | |
| *** erlon_ is now known as erlon | 20:03 | |
| *** erlon has quit IRC | 20:04 | |
| *** erlon has joined #openstack-swift | 20:04 | |
| gholt | peluse: Yessir | 20:11 |
| gholt | I'm waaaay out there | 20:11 |
| creiht | lol | 20:17 |
| peluse | question: looking at a scenario where polices have been partially deployed in a cluster (during upgrade) and the sender (ssync) sends a parm (policy idx) in the msg that the rcvr (not yet upgraded) doesn't understand. The rcv will get an exception and I'm looking to see if there's a way the sender can detect it. Right now it looks like I get a 200 resp and an 'early disconnect' w/no | 20:19 |
| peluse | addtl info | 20:19 |
| peluse | wondering if there's any way for the sender to know more about what happened at the rcvr... | 20:19 |
| *** vu_ has joined #openstack-swift | 20:20 | |
| portante | so that as the cluster rolls over to the new code, new ssync senders start sending to old ssync receiver nodes ... | 20:20 |
| *** vu_ has quit IRC | 20:31 | |
| *** csd has quit IRC | 20:33 | |
| *** kun_huang has quit IRC | 20:34 | |
| gholt | peluse: I'm sure there can be a way, but it might need to be coded up. | 20:35 |
| gholt | I assume you're hitting one of the except Exception catchers and it's not sending anything back to the sender, just dropping the connection. | 20:36 |
| peluse | gholt: thanks, yeah it didn't look like the error handling that's there would lend itself real well to having the sender get details about issues at the rcv other than "didn't work" | 20:36 |
| gholt | Well, there are some error handled, like 408 (timeout) and such. | 20:36 |
| gholt | But no, I didn't handle this future condition I didn't know about. :D | 20:37 |
| peluse | given that ssync isn't "production" yet though, I wonder if we have a legacy case to worry about... is there a timeframe when we'll support ssync as production? | 20:37 |
| peluse | gholt: heh, how could not have predicted this :) | 20:38 |
| gholt | It's production for at least one cluster I know of. And probably 5 more clusters soon enough. | 20:38 |
| gholt | But yeah, it is just me. I can't imagine anybody else is really testing ssync yet. | 20:38 |
| gholt | I personally wouldn't mind if it just didn't work with a kind of unclear reason. I'm pretty sure I could tell what was happening in the logs, right? | 20:39 |
| gholt | On the receiver side that is. | 20:39 |
| peluse | gholt: On the rcvr side you will see an exception for a .split that's complaining about not having the right number of values | 20:40 |
| portante | traceback and all | 20:40 |
| gholt | Yeah, I'm completely fine with that, as the only ssync production user. And I can check for that particular error after some time to see if maybe an upgrade failed on a server. | 20:40 |
| peluse | gholt: so then the sender gets an 'early disconnect' error and moves on. was hoping to isolate that specific error and put something in the log saying "hey this is due to mismatch of swift versions" | 20:40 |
| peluse | gholt: cool, FWIW it is coded to correctly handle the other mismatch where the rcvr has been upgraded but the sender hasn't yet... | 20:41 |
| gholt | Yeah, I'm surprised you don't get back *something* from the first except Exception as err | 20:41 |
| peluse | I posted the err, one sec | 20:42 |
| gholt | It has a yield .... str(err) that should've been sent back to the sender. But if that fails, then it just logs and moves on of course | 20:42 |
| peluse | https://gist.github.com/peluse/9393383 is what you'll see in the logs | 20:42 |
| gholt | That's the sender side right? What's the receiver side log? | 20:43 |
| gholt | Oh that is the receiver, that is a bit confusing. :( | 20:43 |
| peluse | yeah, its my AIO | 20:44 |
| *** vu_ has joined #openstack-swift | 20:44 | |
| gholt | Hah, gotcha. I was just reading the STDOUT exception. didn't see the other one way to the right | 20:44 |
| openstackgerrit | Clay Gerrard proposed a change to openstack/swift: make probetests work with conf.d configs https://review.openstack.org/78753 | 20:45 |
| clayg | ^ yes! negative diff! | 20:45 |
| peluse | so I see the yield ':ERROR: %d %r\n' % (0, str(err)) that you were talking about but I don't see it on the sender side anywhere (how to get at that info)? | 20:45 |
| gholt | peluse: So that definitely means that it sent the error line back to the sender | 20:45 |
| gholt | On that I'm not sure. ;) | 20:46 |
| gholt | Maybe a case of catching the broken pipe error and then attempting to read anything from the socket. | 20:46 |
| gholt | I would definitely consider it extra credit, but it could be useful (esp someday in the future) | 20:47 |
| peluse | gholt: cool, I may mess with it a bit and let you know how it goes | 20:48 |
| clayg | peluse: so torgomatic is working on something similar for the container-replicator - basically teaching the reciever how to read both formats but then having the sender send the *old* format until there's acctually a second (non-default) storage policy | 20:48 |
| clayg | peluse: the idea is that you'll update your cluster to the new storage policy code before you acctually turn on a new storage policy - so by the time you start sending the new format everyone knows how to read it. | 20:49 |
| *** vu_ has quit IRC | 20:49 | |
| gholt | The 2step; always fun. :) | 20:50 |
| gholt | I guess 3step if you include the last part removing the crappy old version of the code. But we seldom do that since we don't know when everybody upgrades... | 20:51 |
| *** csd has joined #openstack-swift | 20:52 | |
| *** vu_ has joined #openstack-swift | 20:52 | |
| peluse | clayg: cool... so the thinking is that the docs will clearly state that you can't start creating new policies until all nodes are udpated. Until then, the issue with ssync is that w/o any kind of API versioning between the two, the sender can't know when its OK to use the new msg format to the rcvr so it just starts sending the new format when the node is updated... | 20:54 |
| peluse | and then the rcvr gets a msg it doens't understand until its updated | 20:54 |
| clayg | peluse: so send the old format | 20:54 |
| peluse | clayg: until when? | 20:55 |
| clayg | peluse: until swift.conf has a storage-policy:1 ??? | 20:55 |
| peluse | clayg: ahh.... | 20:55 |
| clayg | ;) | 20:56 |
| clayg | it was torgomatic's idea | 20:56 |
| * clayg bows down | 20:56 | |
| torgomatic | don't blame me ;) | 20:56 |
| * peluse is not worthy | 20:56 | |
| peluse | torgomatic: are you just checking the global POLICIES are actually doing something with swift.conf? | 20:57 |
| torgomatic | peluse: just checking POLICIES | 20:57 |
| peluse | damn, that's a really good idea... | 20:58 |
| torgomatic | you should have seen the first couple ideas... ask clayg about them; they were pretty bad | 20:58 |
| clayg | that's torgomatic's way of saying the were my ideas | 20:59 |
| peluse | well, I was about ready to code an API versionsing scheme so... :) | 20:59 |
| torgomatic | hehe... please assign credit for good ideas to me, blame for bad ones to /dev/null ;) | 20:59 |
| peluse | hey, unrelated question: I jsut had a jenkins failure on python26 but not python27 and it makes no sense why. Never seen that before and I don't have python26 on my dev AIO. Suggestions? | 21:00 |
| *** jamieh_ has quit IRC | 21:00 | |
| torgomatic | go add the deadsnakes PPA to your system (assuming Ubuntu here, since Cent/RHEL have 2.6) and install python 2.6 | 21:01 |
| torgomatic | tox will pick up on it and you'll be able to test on both | 21:01 |
| openstackgerrit | Clay Gerrard proposed a change to openstack/python-swiftclient: add "info" as an alias to "capabilities" https://review.openstack.org/78710 | 21:02 |
| peluse | torgomatic: thanks,tried that and it hung so I assumed that PPA wasn't alive right now | 21:04 |
| *** vu_ has quit IRC | 21:05 | |
| openstackgerrit | Fabien Boucher proposed a change to openstack/python-swiftclient: TCP port is appended two time in ClientException https://review.openstack.org/78761 | 21:05 |
| peluse | torgomatic: anything wrong with using this function to see if policies have been defined yet (python wise): POLICIES.__len__() | 21:12 |
| torgomatic | you mean len(policies) ? | 21:12 |
| torgomatic | yeah, something like len(policies) == 1 and policies[0].index == 0 | 21:13 |
| peluse | yeah, or that. wasn't sure why the __len__ method existed | 21:13 |
| torgomatic | just to provide support for len() operator | 21:14 |
| torgomatic | operator? function? builtin thingy? whatever it's called... | 21:14 |
| peluse | excellent | 21:14 |
| openstackgerrit | David Goetz proposed a change to openstack/swift: copy over swift.authorize stuff into subrequests https://review.openstack.org/78766 | 21:15 |
| *** NM has joined #openstack-swift | 21:20 | |
| gholt | torgomatic: ^ that just "got" us but dfg fixed it up -- totally missed that as reviewers. :/ | 21:21 |
| torgomatic | gholt: :| | 21:22 |
| * torgomatic should really write some multi-user functional tests for this stuff | 21:22 | |
| gholt | Well that particular one got us but only [ds]los through our cdn layer | 21:23 |
| openstackgerrit | James Carey proposed a change to openstack/python-swiftclient: Remove ignore of H501 hacking check https://review.openstack.org/78772 | 21:25 |
| gholt | I suppose it would also affect anybody using more than one auth | 21:28 |
| clayg | torgomatic: hey if you're having trouble getting -e py26 running on your saio you could always try -> https://github.com/swiftstack/vagrant-swift-all-in-one | 21:30 |
| clayg | ^ also a great way to try out the feature-ec branch since it has baked in support for setting up storage polices | 21:30 |
| clayg | *optional support | 21:31 |
| peluse | torgomatic: FYI you only need to check len(policies) and not the index as its not possible to have just one policy that's not index 0 (if you try to create one policy with idx 1 the policy code will create a policy 0 automatically thus giving you 2 policies) | 21:34 |
| torgomatic | peluse: good to know, thanks | 21:41 |
| *** NM has quit IRC | 21:44 | |
| *** Midnightmyth has quit IRC | 21:48 | |
| clayg | good god why do we add the device and partition to the path during bufferedhttp http_connect? | 21:55 |
| gholt | Some decision we made probably out of convenience 5 years ago. ;) | 21:56 |
| clayg | lol | 21:56 |
| creiht | hah | 21:56 |
| creiht | has it been that long? | 21:56 |
| creiht | aww dang... getting close | 21:56 |
| portante | 711560 | 22:05 |
| torgomatic | That's Numberwang! | 22:06 |
| openstackgerrit | paul luse proposed a change to openstack/swift: Add Storage Policy Support to ssync https://review.openstack.org/65347 | 22:08 |
| peluse | portante: ^ will handle old-->new or new-->old now | 22:10 |
| *** piyush1 has left #openstack-swift | 22:12 | |
| *** erlon has quit IRC | 22:12 | |
| *** tdasilva has left #openstack-swift | 22:39 | |
| *** NM has joined #openstack-swift | 22:56 | |
| *** Midnightmyth has joined #openstack-swift | 23:11 | |
| *** NM has quit IRC | 23:32 | |
| *** dmsimard has quit IRC | 23:39 | |
| *** Midnightmyth has quit IRC | 23:46 | |
| *** fbo is now known as fbo_away | 23:47 | |
| *** changbl has quit IRC | 23:56 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!