| *** kprice789 has quit IRC | 00:14 | |
| openstackgerrit | Merged openstack/swift master: Bump up ceph tests timeout https://review.opendev.org/703848 | 01:11 |
|---|---|---|
| *** gyee has quit IRC | 01:13 | |
| *** ccamacho has quit IRC | 02:50 | |
| openstackgerrit | Thiago da Silva proposed openstack/python-swiftclient master: object versioning features https://review.opendev.org/691877 | 03:06 |
| *** manuvakery has joined #openstack-swift | 03:07 | |
| *** psachin has joined #openstack-swift | 03:35 | |
| openstackgerrit | Thiago da Silva proposed openstack/swift master: Add etag quoter to saio proxy config https://review.opendev.org/704920 | 03:44 |
| openstackgerrit | Thiago da Silva proposed openstack/swift master: Dockerhub description of saio image https://review.opendev.org/704923 | 04:38 |
| openstackgerrit | Thiago da Silva proposed openstack/swift master: Dockerhub description of saio image https://review.opendev.org/704923 | 04:41 |
| *** zaitcev has quit IRC | 04:49 | |
| openstackgerrit | Thiago da Silva proposed openstack/swift master: Add new versioning flag to docker image https://review.opendev.org/704924 | 04:57 |
| *** manuvakery has quit IRC | 05:17 | |
| *** rdejoux has joined #openstack-swift | 07:15 | |
| *** rdejoux has quit IRC | 07:20 | |
| *** tesseract has joined #openstack-swift | 07:47 | |
| *** rdejoux has joined #openstack-swift | 08:05 | |
| *** rpittau|afk is now known as rpittau | 08:14 | |
| *** tkajinam has quit IRC | 08:15 | |
| *** ccamacho has joined #openstack-swift | 08:19 | |
| *** pcaruana has quit IRC | 10:46 | |
| *** rpittau is now known as rpittau|bbl | 11:14 | |
| *** csmart has left #openstack-swift | 11:30 | |
| *** csmart has quit IRC | 11:30 | |
| *** pcaruana has joined #openstack-swift | 11:48 | |
| *** jvisser has joined #openstack-swift | 12:16 | |
| *** pcaruana has quit IRC | 12:21 | |
| *** mvkr has quit IRC | 12:26 | |
| *** jvisser has quit IRC | 12:27 | |
| *** jvisser has joined #openstack-swift | 12:27 | |
| *** jvisser has quit IRC | 13:19 | |
| *** rpittau|bbl is now known as rpittau | 13:29 | |
| *** jvisser has joined #openstack-swift | 13:51 | |
| *** pcaruana has joined #openstack-swift | 14:03 | |
| *** psachin has quit IRC | 14:32 | |
| *** pcaruana has quit IRC | 14:34 | |
| *** pcaruana has joined #openstack-swift | 15:02 | |
| *** ccamacho has quit IRC | 15:27 | |
| *** psachin has joined #openstack-swift | 15:29 | |
| *** camelCaser has quit IRC | 15:49 | |
| *** camelCaser has joined #openstack-swift | 15:53 | |
| *** dsariel has joined #openstack-swift | 16:03 | |
| *** gyee has joined #openstack-swift | 16:04 | |
| *** jvisser_ has joined #openstack-swift | 16:07 | |
| *** jvisser_ has quit IRC | 16:08 | |
| *** mvkr has joined #openstack-swift | 16:13 | |
| *** rdejoux has quit IRC | 16:44 | |
| openstackgerrit | David Sariel proposed openstack/swift master: DNM - dd probe tests https://review.opendev.org/705063 | 16:50 |
| *** tesseract has quit IRC | 17:08 | |
| *** dsariel has quit IRC | 17:10 | |
| *** evrardjp has joined #openstack-swift | 17:33 | |
| *** rpittau is now known as rpittau|afk | 18:02 | |
| timburke | clayg, in or out for 2.24? p 688594 | 18:04 |
| patchbot | https://review.opendev.org/#/c/688594/ - swift - py38: cgi lost some names - 6 patch sets | 18:04 |
| rledisez | I'm working the etag_hasher middleware that calculate md5 for every object PUT (in place of crypto, or proxy EC). I'm at a point where, without the middleware, the proxy-server on its is not working because it want to store the client-facing etag in a meta (X-Object-Sysmeta-EC-Etag). It feels wrong that without a middleware (even mandatory / auto-inserted) the core does not work. should I allow to store an object without client- | 19:28 |
| rledisez | facing etag? it also feels wrong… | 19:28 |
| clayg | rledisez: yeah that's difficult! we *have* extracted proxy behavior out to required/auto-inserted middleware as an implementation detail in the past | 20:02 |
| clayg | but I'm not sure it was ever... "required" ... I mean sysmeta is pretty broken w/o gatekeeper! | 20:02 |
| clayg | rledisez: where in the pipeline will the hasher go? | 20:03 |
| rledisez | clayg: I would say right before crypto, so to the right, between the mandatory slo/dlo/copy and proxy-logging | 20:06 |
| rledisez | it would look like | 20:07 |
| rledisez | copy slo dlo etag_hasher [cryptoØ proxy-logging proxy-server | 20:07 |
| clayg | yup yup yup | 20:07 |
| clayg | I don't know... i guess I would probably make the proxy robust (where it doesn't blow up) to if someone does that trick where you put in a noop middleware with the same name to prevent the auto-insert | 20:08 |
| rledisez | clayg: in case an operator doesn't care about MD5 because its application does not need it. yeah, I can see a use-case here | 20:10 |
| clayg | but maybe even that's not so easy :'( | 20:10 |
| rledisez | but, auditor won't find an etag in case of replication policy… | 20:10 |
| clayg | yeah i'm not sure it's a use-case we care about THAT much, maybe if someday it helps using other CRCs than md5 | 20:11 |
| *** zaitcev has joined #openstack-swift | 20:11 | |
| *** ChanServ sets mode: +v zaitcev | 20:11 | |
| clayg | rledisez: yeah I guess it's like the early work to making things not assume they always have the calculated & verified md5 written in metadata | 20:12 |
| clayg | and then thinking through the implications of a failure mode that's basically "no etag => oh well, verify less" | 20:12 |
| rledisez | i'm thinking it could only work if the separation between client-facing and cluster-facing checksum is done (which is my target, so that cluster-facing can be something else than MD5) | 20:13 |
| clayg | maybe the "default" the proxy writes down when the hasher didn't provide the client facing etag is some sort of sentinal like "algo: null; value: you are screwed" | 20:13 |
| clayg | as opposed to just "no etag found; i'm sure it's fine!" | 20:13 |
| clayg | rledisez: yeah that's awsome!!! there is definately a use-case for clients that want different checksum done too - and letting operators have choice on the backend is a super big win | 20:14 |
| clayg | this sounds GREAT!!! | 20:14 |
| rledisez | clayg: thx for thinking with me, I think i see clearer now :) | 20:15 |
| *** psachin has quit IRC | 20:16 | |
| openstackgerrit | Thiago da Silva proposed openstack/swift master: Dockerhub description of saio image https://review.opendev.org/704923 | 20:17 |
| openstackgerrit | Merged openstack/swift master: py38: cgi lost some names https://review.opendev.org/688594 | 20:32 |
| openstackgerrit | Merged openstack/swift master: Add py38 unit test job https://review.opendev.org/689599 | 20:32 |
| *** rcernin has quit IRC | 20:36 | |
| *** jvisser has quit IRC | 20:50 | |
| *** rdejoux has joined #openstack-swift | 21:14 | |
| *** rdejoux has quit IRC | 21:18 | |
| *** rdejoux has joined #openstack-swift | 21:19 | |
| *** jvisser has joined #openstack-swift | 21:23 | |
| *** rdejoux has quit IRC | 21:24 | |
| *** jvisser has quit IRC | 21:28 | |
| *** jvisser has joined #openstack-swift | 21:28 | |
| openstackgerrit | Thiago da Silva proposed openstack/swift master: Add etag quoter to saio proxy config https://review.opendev.org/704920 | 21:47 |
| rledisez | one question about crypto middleware. what's the reason for the encrypter middleware to calculate the MD5 of encrypted data. it feels like it's only to ensure there is no data corruption after the middleware (on the wire?). As it's not always done when crypto is not enabled, I'm enclined to think it should be removed. did I miss something? | 21:48 |
| tdasilva | rledisez: isn't that the md5 that gets stored in the etag header? IIRC it was needed for auditor (or any other processes that rely on the etag) | 22:02 |
| rledisez | tdasilva: yes but in replica it's calculated by the obj-server. with EC it's calculated by the proxy-server (with the MD5 of the frags). | 22:05 |
| rledisez | so, it is used as a validation that the right data was written on the obj-server. but we don't do that for (eg) replica | 22:06 |
| *** jvisser has quit IRC | 22:21 | |
| mattoliverau | morning | 22:24 |
| tdasilva | rledisez: yeah, looking quickly through code couldn't come up with a better reason than what you said... | 22:55 |
| tdasilva | couldn't find much in docs either | 22:57 |
| *** paladox is now known as paladox|UKInEU | 23:00 | |
| *** tkajinam has joined #openstack-swift | 23:10 | |
| *** rcernin has joined #openstack-swift | 23:20 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!