openstackgerrit | Takashi Kajinami proposed openstack/storlets: WIP: Refactor daemon factory https://review.openstack.org/343687 | 03:13 |
---|---|---|
*** cdoron has joined #openstack-storlets | 09:39 | |
*** cdoron has quit IRC | 12:59 | |
*** cdoron has joined #openstack-storlets | 13:00 | |
eranrom_ | Hi | 13:01 |
kota_ | hi | 13:01 |
eranrom_ | hey. | 13:01 |
cdoron | hi | 13:01 |
eranrom_ | hey Doron\ | 13:01 |
eranrom_ | a | 13:01 |
eranrom_ | kota_: How was the Hackathon? | 13:01 |
kota_ | eranrom: that was so great. | 13:02 |
kota_ | we had a lot of progressive conversation there. | 13:02 |
eranrom_ | kota_: Any new companies involved? | 13:03 |
kota_ | ah, i don't think new companies there but I met some of new contributors | 13:04 |
eranrom_ | kota_: nice! Do you know if Takashi will join today? | 13:04 |
kota_ | almostly rackers (including OSIC) and guys from IBM, maybe? | 13:05 |
eranrom_ | kota_: I guess the go stuff brought some new faces... | 13:05 |
kota_ | he said he might be absent or delayed. but never mind to start w/o him. | 13:05 |
eranrom_ | ok. | 13:05 |
eranrom_ | So Doron joined us for the beginning of the discussion, so would like to start with his topic. | 13:06 |
eranrom_ | Doron? | 13:06 |
cdoron | hi all | 13:06 |
kota_ | great | 13:06 |
cdoron | I have recently experimented with a script for setting up the storlet engine on DEVSTACK | 13:07 |
cdoron | using vagrant | 13:07 |
cdoron | Which got me thinking about whether we can use this, or something similiar, in the storlets project | 13:07 |
cdoron | One question could be, whether we want to continue running the storlet tests on an all-in-on Swift | 13:08 |
kota_ | interesting | 13:08 |
cdoron | or do we want to replace all-in-one with devstack? | 13:08 |
cdoron | Another question (orthogonal question): | 13:08 |
cdoron | Do we want to run the test against a swift installation on a docker container | 13:09 |
cdoron | rather than on the all-in-one | 13:09 |
cdoron | (in the docker container we could either have devstack | 13:09 |
cdoron | or an all-in-one swift) | 13:09 |
cdoron | Thoughts? | 13:09 |
kota_ | so can I make sure the mean of all-in-one? | 13:10 |
cdoron | I mean the current swift installation | 13:10 |
kota_ | you mean vagrant-swift-all-in-one or an instance result from s2io setup? | 13:10 |
cdoron | which uses the scripts to install on the current machine | 13:10 |
cdoron | I meant the s2io setup | 13:11 |
kota_ | ok | 13:11 |
cdoron | (I am not sure we can run vagrant on a VM, so let us forget about vagrant for the moment) | 13:11 |
cdoron | If we have a script for installing swift+storlet-engine on a docker container, people could use it to easily experiment with the storlet engine | 13:13 |
cdoron | It is less intrusive than installing swift on a regular machine | 13:13 |
eranrom_ | So I think we have two orthogonal questions which can be translated to perhaps 4 ways to get an s2aio like deployment: | 13:14 |
eranrom_ | {docker, vm} X {devstack, current swift ansbile install} | 13:14 |
cdoron | vm == jenkins vm ? | 13:15 |
eranrom_ | yes | 13:15 |
eranrom_ | well, or a dev end that a developer sets for himself | 13:15 |
eranrom_ | end=env | 13:15 |
eranrom_ | I am assuming all 4 can be tested from jenkins | 13:16 |
eranrom_ | kota_: do you know whether Swift today has any jenkins / gate jobs that rely on something else form SAIO? | 13:17 |
eranrom_ | e.g. devstack? | 13:18 |
kota_ | ah | 13:18 |
kota_ | if we can use devstack template, it's easy to add the gate | 13:18 |
kota_ | probably, project-config has the template script for that. | 13:19 |
kota_ | and for SAIO thing, swiftstack sets a job for non-voting gate which is third party. | 13:19 |
kota_ | I an not familier with such a third party gate though. | 13:20 |
eranrom_ | kota_: When I am doing a 'git review' do you know which template is used for the jenkins test jobs | 13:20 |
kota_ | I think the template devstack will complete basic (whole?) component for any openstack projects (not sure, it might be only core project) | 13:20 |
eranrom_ | So it seems a good idea to have that as well. | 13:21 |
eranrom_ | Esp if we want to be an official project | 13:21 |
kota_ | eranrom: depends on project config. | 13:21 |
kota_ | wait...looking at project-config | 13:21 |
eranrom_ | Well, I guess that we can change (with infra help) the project config so that we test both against devstack template and against our current 's2aio setup' which uses 'clean' ubuntu.14_04 | 13:22 |
eranrom_ | now waiting :-) | 13:22 |
-kota_- - gate-storlets-pep8 | 13:23 | |
-kota_- - gate-storlets-python27 | 13:23 | |
-kota_- - gate-storlets-functional | 13:23 | |
kota_ | you know :) | 13:23 |
eranrom_ | actually, Doron was the one who did that initial setup... | 13:24 |
kota_ | iirc, gate-(project name)-pep8 will run tox -e pep8 | 13:24 |
kota_ | as well as it, gate-(project)-devstack-something can be available for the gates. | 13:25 |
kota_ | most likely, | 13:25 |
kota_ | - gate-swift-dsvm-functional | 13:25 |
kota_ | dsvm means devstack-virtual machine, i think | 13:26 |
eranrom_ | ok! so looks like (1) this can be done, and (2) this is a requirement from all projects. | 13:26 |
kota_ | and -functional will kick .functests probably. | 13:26 |
eranrom_ | kota_: make sense. | 13:27 |
eranrom_ | this is also what I think should be kicked for storlets | 13:27 |
eranrom_ | anything else on this subject? | 13:30 |
kota_ | I also think if we can run storlets with devstack, it seems great. On the other hand, s2io also be helpful for newbees to deploy not-all-in-one environment. | 13:30 |
kota_ | s2aio | 13:31 |
eranrom_ | kota_: Agree. In fact I was thinking of having both. | 13:31 |
eranrom_ | kota_: If this is too much for each patch, perhaps we can have the devstack one as part of a pre-release test. | 13:31 |
eranrom_ | I mean, I think that there are such tests, but need to find out... | 13:32 |
kota_ | yup | 13:32 |
kota_ | and note that the gate devstack always runs on the master, afaik | 13:32 |
eranrom_ | right, which is currently a problem for us :-) this brings to the next topic, unless there is more on this one. | 13:33 |
kota_ | need to make sure though. it might be broken affected from any other openstack project changes. | 13:33 |
kota_ | particuler Swift/Keystone. | 13:33 |
kota_ | k | 13:33 |
eranrom_ | right. good point | 13:33 |
eranrom_ | Doron? | 13:34 |
eranrom_ | ok, Doron needed to drop. | 13:35 |
eranrom_ | next topic is a release. | 13:35 |
eranrom_ | At the time Takashi and I thought that we would do 3 things before 'cutting' a Storlets mitaka release. | 13:36 |
eranrom_ | 1. land the range patch - done. | 13:36 |
kota_ | nice | 13:36 |
eranrom_ | 2. land the storlet docker request refactoring - done. | 13:36 |
eranrom_ | 3. move the installation and test cod to work with Keystone V3 API | 13:37 |
eranrom_ | not done yet | 13:37 |
kota_ | wooo, v3 :/ | 13:37 |
eranrom_ | problem? | 13:37 |
kota_ | no? | 13:38 |
kota_ | just I'm a bad guy for keystone :P | 13:38 |
eranrom_ | I see :-) | 13:39 |
eranrom_ | Anyway, I am basically ok with this (although I am not sure yet what needs to be done), is that I think the Spark/Hadoop echo system currently does not support V3 auth | 13:40 |
eranrom_ | so I would | 13:40 |
eranrom_ | not want to break their ability to work with storlets. | 13:40 |
eranrom_ | That is unless a Keystone server can be used for both v2 and v3 auth at the same time. | 13:41 |
kota_ | yup | 13:41 |
eranrom_ | Does being a bad guy with Keystone means you would not know :-) | 13:41 |
kota_ | Or doing with frustrating :-) | 13:43 |
eranrom_ | I see. | 13:43 |
eranrom_ | :-) | 13:43 |
eranrom_ | Perfectly understood! | 13:43 |
eranrom_ | Its never been easy with Keystone | 13:43 |
kota_ | eranrom: you konw ;) | 13:43 |
eranrom_ | ok, I will try to find out if it can support both. If not I would recommend to have a release with V2 | 13:45 |
eranrom_ | So that we will have something spark/hadoop can work with | 13:45 |
kota_ | sounds reasonable | 13:45 |
eranrom_ | ok, this is all from me | 13:45 |
kota_ | from my side... not so a big topic. | 13:46 |
kota_ | takashi is still working on something | 13:46 |
kota_ | like... | 13:46 |
kota_ | thought for copy middleware placemnet | 13:47 |
kota_ | in the pipeline | 13:47 |
kota_ | and consideration for python supports. | 13:47 |
eranrom_ | right | 13:47 |
kota_ | I think nothing special for those right now. | 13:47 |
eranrom_ | ok. | 13:48 |
kota_ | probably, next week he could share thought for them | 13:48 |
eranrom_ | ok sure. So I guess this is all for today | 13:48 |
eranrom_ | Thanks very much for joining | 13:48 |
kota_ | ah, I forgot a special for today, say "thank you eronrom for submitting proposal for barcelona" | 13:49 |
eranrom_ | :-) | 13:49 |
kota_ | :-) | 13:49 |
eranrom_ | good luck to us. | 13:49 |
* kota_ hope so | 13:49 | |
kota_ | ah, sorry, one more thing | 13:50 |
kota_ | next week meeting is the same time as today? | 13:50 |
kota_ | or will do at a new schedule? (takashi concerned if you could give a feedback for the schedule) | 13:51 |
kota_ | takashi is available 2 pm or 4 pm on Tsu in JST (the first proposal is 3 pm iirc) | 13:52 |
kota_ | anyway, not urgnet. we could define the schedule later though. | 13:53 |
*** cdoron has quit IRC | 14:09 | |
eranrom_ | kota_: sorry seen this just now. Takashi confirmed he can do Tue 08:00 UTC. | 14:51 |
eranrom_ | If this is ok I will change the time on the wiki. | 14:51 |
eranrom_ | Also, I will be on vacation next week. Coming back Sunday the week after (July 31st) | 14:53 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!