| *** mattoliverau has quit IRC | 00:00 | |
| *** matt6434 has joined #openstack-storlets | 00:12 | |
| *** matt6434 has quit IRC | 00:17 | |
| *** matt6434 has joined #openstack-storlets | 00:25 | |
| *** matt6434 is now known as mattoliverau | 00:27 | |
| *** openstack has joined #openstack-storlets | 00:53 | |
| *** openstack has joined #openstack-storlets | 09:08 | |
| *** openstackgerrit has quit IRC | 09:32 | |
| *** openstackgerrit has joined #openstack-storlets | 09:33 | |
| *** openstackgerrit has quit IRC | 10:02 | |
| *** openstackgerrit has joined #openstack-storlets | 10:03 | |
| *** [1]eranrom has joined #openstack-storlets | 11:37 | |
| *** eranrom has quit IRC | 11:37 | |
| *** [1]eranrom is now known as eranrom | 11:37 | |
| *** eranrom has quit IRC | 11:42 | |
| *** takashi has joined #openstack-storlets | 12:02 | |
| *** takashi has quit IRC | 13:49 | |
| *** takashi has joined #openstack-storlets | 13:49 | |
| *** eranrom has joined #openstack-storlets | 13:50 | |
| kota_ | hello | 13:57 |
|---|---|---|
| kota_ | meeting will start in 3 minutes here, right? | 13:57 |
| takashi | Hi | 13:58 |
| takashi | kota_: I suppose so | 13:59 |
| eranrom | Hi | 13:59 |
| eranrom | Who is on? | 13:59 |
| takashi | I am | 13:59 |
| hrou | here ! | 14:00 |
| eranrom | kota_,? | 14:00 |
| kota_ | sure | 14:00 |
| eranrom | peluse,? | 14:01 |
| eranrom | Doron told me he can't make it. | 14:01 |
| eranrom | I have updated the agenda: https://wiki.openstack.org/wiki/Meetings/Storlets | 14:01 |
| eranrom | topic: NTT Updates :-) | 14:02 |
| eranrom | takashi, stage is yours. | 14:02 |
| takashi | eranrom: OK | 14:02 |
| takashi | As I told eranrom before, we tested Storlet last month. | 14:03 |
| takashi | We tested Storlets with Swift Liberty(2.3.0). | 14:04 |
| kota_ | liberty is 2.5.0, fyi | 14:04 |
| takashi | kota_: Thanks, 2.5.0 | 14:04 |
| hrou | takashi ! That's great did you run into some failures ? (expected as storlets on the object server do not work as of 2.4 or what not) | 14:05 |
| takashi | hrou: yes, we have to fix proxy-server to run storlet. | 14:06 |
| takashi | After fixing that, I could run storlets with Liberty. | 14:06 |
| takashi | I'm going to commit that fix of proxy-server soon. | 14:06 |
| hrou | takashi, this fix allows storlets to run on the object-server ? | 14:06 |
| kota_ | hrou: i think so | 14:07 |
| takashi | hrou: Yes | 14:07 |
| kota_ | to get start to catch up of storlets status | 14:07 |
| hrou | Oh wow ! Very curious to see that fix. | 14:07 |
| kota_ | at fist, we ran that in swift-all-in-one | 14:07 |
| kota_ | environment | 14:07 |
| takashi | using scripts included in storlet repository | 14:08 |
| kota_ | and, after that, he set up cluseter environment | 14:08 |
| takashi | we are now running environment with 2 proxy-servers, 3 storage-servers, and 1 keystone | 14:08 |
| kota_ | for running on the cluster environment, he have to fix some issues | 14:08 |
| kota_ | s/have to/had to/ | 14:09 |
| hrou | Great! | 14:09 |
| eranrom | This is great! | 14:09 |
| kota_ | and with his fixes, currently storlets seems to work well completely | 14:09 |
| kota_ | that's awesome. | 14:09 |
| eranrom | ABSOLUTELY!!! | 14:09 |
| takashi | kota_: thank you for your following up | 14:09 |
| hrou | +100 :) | 14:10 |
| eranrom | Do you want to elaborate on the fixes or just sent pathces? | 14:10 |
| eranrom | s/sent/send | 14:10 |
| takashi | I think I can send the patch soon. I already finished creating some tests for that. | 14:11 |
| eranrom | Cool! Looking forward | 14:11 |
| hrou | Yea very interested in seeing the swift patch too ! | 14:12 |
| eranrom | Can you tell us a little on your demo app? | 14:12 |
| kota_ | and now, we are working for listing what we can be fixed and what can be improved in future work. | 14:13 |
| eranrom | Great! | 14:13 |
| kota_ | eranrom: maybe, takashi? | 14:13 |
| takashi | eranrom, kota_: I created some image processing app, like feature detection in images. | 14:13 |
| eranrom | Nice! | 14:14 |
| takashi | We think image processing is a good starting point, and planning some PoC using that kind of apps | 14:15 |
| eranrom | takashi: I assume you are invoking the Storlet on GET - right? | 14:16 |
| takashi | eranrom: Yes | 14:16 |
| takashi | We store raw images in Swift, and invoke storlet on GET, and get a result of processing. | 14:17 |
| eranrom | takashi: gotcha. Just out of curiousity, I assume you need some non trivial dependencies are these part of the Storlet or are you using the dependency mechanism or deploy a Docker image that has it? | 14:18 |
| eranrom | I mean the image processing engine | 14:18 |
| takashi | eranrom: We put the engine in the Docker container image. We are now using OpenCV, which needs so many files that it is very difficult to add them using dependency mechanism... | 14:19 |
| eranrom | takashi: This is great, did you use the ansible 'managememnt' scripts we have? | 14:20 |
| cdoron | hi there. Sorry I'm late | 14:21 |
| takashi | eranrom: Yes, I used them. | 14:21 |
| eranrom | cdoron: Hi! | 14:22 |
| hrou | hey cdoron ! | 14:22 |
| eranrom | takashi: Now I am even more excited! Did you run into issues there as well (I assume yes :-) | 14:22 |
| takashi | eranrom: I didn't met so big issues. I think the biggest problem was the one about ssh key management. It worked well for me. | 14:23 |
| eranrom | YEY!!! :-) | 14:24 |
| hrou | That's really great to hear, I think the scripts are much improved now from the early days :) | 14:24 |
| eranrom | So I am really lokking forward to the patches and your ideas for future work! | 14:24 |
| hrou | Yep the SSH key management is usually the key | 14:24 |
| takashi | hrou: definitly | 14:25 |
| takashi | eranrom: So, I'm going to commit patch for swift, which we need to run Storlet, | 14:26 |
| kota_ | on master, right? | 14:26 |
| takashi | and also going to commit some fixes or improvements about Storlet after creating a list | 14:26 |
| takashi | kota_: Yes, on master | 14:26 |
| eranrom | takashi, That's great. BTW the Swift patch is it big? | 14:26 |
| takashi | eranrom: It's not so big | 14:26 |
| hrou | You mentioned proxy server - is it contained to that ? | 14:27 |
| takashi | hrou: yes | 14:27 |
| cdoron | takashi: a while ago you mentioned unit-tests. Which components are you interested in testing? | 14:27 |
| hrou | takashi, great ! I did talk to Sam about it back in Austin and he said as long as it wasn't huge the community may be Ok with it; It was the EC refactoring that introduced the issue ;) | 14:28 |
| takashi | cdoron: storlet_handler middleware | 14:28 |
| eranrom | do we have more updates or should we move to the next topic? | 14:30 |
| eranrom | which has to do with basic administrations? | 14:30 |
| takashi | eranrom: I think that' all. | 14:30 |
| eranrom | s/?// | 14:30 |
| eranrom | takshi: Thanks a lot!!! You have made my day :-) | 14:31 |
| eranrom | so the question I wanted to ask has to do with how to manage feature design. | 14:31 |
| eranrom | For example, I really want to start working on COPY (which kota_ asked about in the summit) | 14:32 |
| hrou | Well and what features would be best to focus on : ) | 14:32 |
| eranrom | absolutely. | 14:32 |
| eranrom | Ideas? | 14:32 |
| hrou | eranrom, kota_ can you have a use case you envision for copy ? | 14:32 |
| eranrom | hrou: Generate thumbnail of an existing photo | 14:33 |
| kota_ | i thinks we can manage some processing applicatiton with some times... | 14:33 |
| kota_ | like get some results and store it into swift | 14:33 |
| hrou | eranrom, and in this scenario you want the thumbnail to persist in object storage, otherwise you could do this on get but its a lot more processing | 14:33 |
| kota_ | afiter that, reuse the result in another apps. | 14:34 |
| kota_ | hrou: yup, like that. | 14:34 |
| kota_ | that's an use case, i thought. | 14:34 |
| hrou | Neat yea, I could see that use case. | 14:35 |
| eranrom | kota_: Right, I think that this generalizes to the ability to do things more "asynchronously." | 14:35 |
| hrou | eranrom, which is interesting; akin to Amazon Lambda | 14:35 |
| eranrom | hrou,: sort of. | 14:35 |
| hrou | Well someone still needs to issue that copy (whether it be middleware or ..) but .. | 14:35 |
| eranrom | yes | 14:35 |
| eranrom | right, but that someone does not need to wait for the response and do something with it. | 14:36 |
| eranrom | as the 'response' goes to the srtore | 14:36 |
| hrou | yep! | 14:36 |
| eranrom | well perhaps "asynchronously" is not the right way to describe | 14:37 |
| hrou | Yea I think it'll cause some confusion, but I see why you call it that :) In comparison to the get/put codepath we generalyl describe. | 14:37 |
| kota_ | i think, we can start for the usecase with synchonously | 14:38 |
| kota_ | becaouse it may emproy asyncronous process in out of swift. | 14:38 |
| kota_ | s/emproy/employ | 14:38 |
| kota_ | asynchronous | 14:38 |
| kota_ | a tons of typos... sorry :/ | 14:38 |
| eranrom | hrou, kota_: Agree. I should not have used async here :-) | 14:39 |
| eranrom | So anyway do we want blueprints - which are already there or do something like Swift with a dedicated Git repo for design docs? | 14:39 |
| eranrom | kota_: at 11PM my time I have more | 14:39 |
| hrou | eranrom, yea that's a good question | 14:39 |
| eranrom | kota_: That is typoes :-) | 14:40 |
| hrou | it seems like swift folks are kinda moving away from swiftspec ? Or only for the very large projects it seems | 14:40 |
| kota_ | no idea, either seems to work for us :) | 14:40 |
| hrou | hehe, yea I kinda like swiftspecs to be honest | 14:40 |
| hrou | Find its useful to colleberate on a high level design that way. | 14:40 |
| eranrom | Do you think its a headache to set it up? | 14:40 |
| eranrom | do we need openstack infra guys for that? | 14:41 |
| eranrom | I mean something like swiftspec | 14:41 |
| hrou | Maybe just to register the project yea but its probably pretty minor, really its just any REPO but just with docs | 14:41 |
| hrou | With blueprints ? You can't really comment inline can you ? | 14:41 |
| hrou | Its more like you just add comments like with a bugreport or what not. | 14:42 |
| eranrom | hrou,: Honestly, I do not know. | 14:42 |
| eranrom | So how about starting with an 'unregistered' repo | 14:42 |
| hrou | Yep I think that works well too; | 14:42 |
| kota_ | where do you think to set that? | 14:42 |
| hrou | kota_, what does swift3 do ? | 14:42 |
| kota_ | ah... no official docs in future work, right now :P | 14:43 |
| kota_ | s/in/for/ | 14:43 |
| eranrom | kota_,: I think IBM has a Github 'namespace' we can put it there for a start. | 14:43 |
| kota_ | just a talk with IRC and some online conversation... | 14:44 |
| kota_ | but | 14:44 |
| hrou | kota_, haha :) Gotcha! Yea probably just IRC conversations are generally sufficent. | 14:44 |
| kota_ | it has both pros and cons. | 14:44 |
| kota_ | i sometimes am asked 'what's going on?' from some users | 14:45 |
| kota_ | that's cons :/ | 14:45 |
| eranrom | I can buy starting off with IRC | 14:46 |
| kota_ | but yeah, in flight conversation is so quick to decide some plan for future work. | 14:46 |
| eranrom | well, we do have a log (thanks to hrou) | 14:46 |
| hrou | For anything involved design wise I could see it be a bit nice to have doc we can comment on | 14:47 |
| kota_ | hrou: agree | 14:47 |
| eranrom | ok so lets do both. If someone is interested in a doc, (s)he can put it in a repo we will setup | 14:47 |
| hrou | So eranrom I think we could try the repo route if it isn't too bad, lets go ahead with that, if we run into issues then we can do just IRC for now | 14:47 |
| hrou | eranrom, yep exactly :0 | 14:48 |
| hrou | :) | 14:48 |
| eranrom | ok then | 14:48 |
| eranrom | next topic? | 14:48 |
| kota_ | k | 14:48 |
| takashi | yes | 14:48 |
| eranrom | Bug / Feature prioritization | 14:48 |
| eranrom | so the first item there is what takashi has a fix for. Yey! | 14:49 |
| eranrom | We have discussed copy | 14:49 |
| eranrom | and there is the list that Takashi and Kota will send | 14:49 |
| eranrom | anything else at this point? | 14:50 |
| takashi | I have nothing more this time. | 14:50 |
| eranrom | So shell we end our first IRC meeting with the famous words? | 14:51 |
| eranrom | Thank you for working on Storlets! | 14:51 |
| kota_ | yey! | 14:51 |
| hrou | Thanks all ! | 14:52 |
| hrou | We can use the wiki page for the agenda going forward if anyone wants to discuss something as well | 14:52 |
| takashi | thanks! | 14:52 |
| kota_ | thanks eranrom for organize this :-) | 14:52 |
| eranrom | Thank you! | 14:53 |
| cdoron | thanks | 14:53 |
| takashi | a small thing | 14:55 |
| takashi | I updated my Storlet many times in my multi-cluster env, but faced no updating problem. | 14:55 |
| takashi | eranrom: I think we have to examine more about the situation to cause this problem. https://bugs.launchpad.net/storlets/+bug/1495600 | 14:57 |
| openstack | Launchpad bug 1495600 in storlets "Problem in Storlet cache / container updates" [Undecided,New] | 14:57 |
| eranrom | takashi, ok thanks! Having a second look at it. | 14:59 |
| takashi | eranrom: I'm going to check my logs about that tommorow. | 15:00 |
| eranrom | takashi: ok thanks! | 15:00 |
| takashi | now "today" in JST | 15:00 |
| eranrom | :-) | 15:00 |
| hrou | Actually that's an interesting one; I recall running into that in the early days and as a result had to delete the cache (i.e. the directory containing) to trigger it to update (by fetching it again) | 15:01 |
| hrou | I'll check | 15:01 |
| eranrom | One possibility that comes in to mind is that the same object was always used so somehow the Storlet was executed on the same node all the time, and there might have been other nodes where the Storlet did not get updated. | 15:02 |
| eranrom | I agree that there is a small chance for that though | 15:02 |
| hrou | Do we use x-newest on that ? | 15:02 |
| eranrom | I do not think so | 15:03 |
| eranrom | you mean when copying the Storlet from Swift to the node where it runs>? | 15:03 |
| hrou | Yep | 15:04 |
| eranrom | looknig | 15:04 |
| hrou | Well that's just a GET request right using the IC | 15:04 |
| takashi | hrou: I found that code now. | 15:07 |
| eranrom | right. resp = client.make_request('GET', path, {'PATH_INFO': path}, [200]) | 15:08 |
| eranrom | no x-newest | 15:08 |
| hrou | Great takashi, eranrom ! Yea so maybe x-newest would help a bit | 15:08 |
| hrou | I mean given this is swift, this is all really ok : ) expected in an eventually consistent system | 15:08 |
| eranrom | I quoted the wrong code | 15:13 |
| eranrom | sec. | 15:13 |
| eranrom | So we actually HEAD the storlet object to make sure the user has read access to it - this is essentially how we enforce storlet execution: | 15:15 |
| eranrom | user has read access to storlet container == user can invoke it | 15:15 |
| eranrom | Anyway that request does not have the x-newest | 15:16 |
| eranrom | which is definately a problem | 15:16 |
| eranrom | good catch! | 15:16 |
| hrou | I think after the head we compared the timestamp and make the decision yea. | 15:17 |
| eranrom | right | 15:17 |
| eranrom | Let me update the but with this thought. | 15:17 |
| eranrom | or do you want to? | 15:18 |
| hrou | Sure ! I can fix that, but again it's not a bug per say : ) But we may as well | 15:19 |
| hrou | You had a thought though ? | 15:19 |
| eranrom | I did not mean to actually go fix that now, just update the bug in launchpad | 15:20 |
| hrou | Oh !! :) Sure will do | 15:21 |
| eranrom | even if this is not a bug, it did affect the development work. I know yosef has spent a lot of time chasing why the Storlet is not updated. | 15:21 |
| eranrom | Thanks! | 15:21 |
| eranrom | ok, gotta go. ttyl | 15:21 |
| hrou | Sure thing, thanks eranrom will do ! | 15:22 |
| eranrom | thank you! | 15:22 |
| *** takashi has quit IRC | 15:24 | |
| *** eranrom has quit IRC | 15:26 | |
| *** eranrom has joined #openstack-storlets | 19:20 | |
| *** eranrom has quit IRC | 19:32 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!