| *** dshulyak has joined #openstack-solar | 00:02 | |
| *** dshulyak has quit IRC | 00:07 | |
| *** salmon_ has quit IRC | 02:01 | |
| *** dshulyak has joined #openstack-solar | 03:38 | |
| *** dshulyak has quit IRC | 03:43 | |
| *** dshulyak has joined #openstack-solar | 05:27 | |
| *** dshulyak has quit IRC | 05:31 | |
| *** dshulyak has joined #openstack-solar | 07:10 | |
| *** salmon_ has joined #openstack-solar | 08:30 | |
| bogdando | folks, review please https://review.openstack.org/#/c/286730/ We have this annoying issue since ages. We should keep that w/a unless all resources fixed to not hardcode virtualbox keys paths :) | 08:49 |
|---|---|---|
| bogdando | I think that'd be better than switching to password based auth for the vagrant libvirt provider case | 08:49 |
| bogdando | salmon_, pigmej ^^ | 08:50 |
| openstackgerrit | Maciej Kwiek proposed openstack/solar: Change PENDING_RETRY to ERROR_RETRY in soft_stop https://review.openstack.org/287044 | 08:53 |
| salmon_ | dshulyak: https://github.com/Mirantis/f2s/pull/16 | 09:13 |
| salmon_ | bogdando: do we need to have it in Vagrantfile? | 09:13 |
| pigmej | salmon_: where would you like to have it instead ? | 09:14 |
| bogdando | salmon_, not much choice | 09:14 |
| salmon_ | :( | 09:14 |
| bogdando | or let's fix all of the hardcoded virtualbox paths in all examples | 09:14 |
| bogdando | I hope we will, eventually. then we can remove the w/a from Vagrantfile | 09:15 |
| dshulyak | salmon_: btw i noticed that genkeys are missing in created tasks, i think thats the reason of rsync error | 09:16 |
| salmon_ | dshulyak: to have genkeys task you need to run fsclient master and fsclient alloc 1 master | 09:17 |
| dshulyak | fsclient alloc 1 master ? oO i didnt know that it will work at all :) | 09:17 |
| salmon_ | dshulyak: I tried and it worked. Keys were generated. now there is a problem that in astute they call mkdir -p and we in naive_sync don't | 09:19 |
| dshulyak | yeah, i was creating manually those directories | 09:20 |
| salmon_ | pigmej: will add it to rsync handler :) | 09:20 |
| pigmej | ? | 09:21 |
| salmon_ | mkdir -p :) | 09:21 |
| pigmej | ah, Yeah, I will extend our rsync handler to "create required folders" | 09:22 |
| mkwiek | Hi guys, I have a question about this bug: https://bugs.launchpad.net/solar/+bug/1544537 it's about making transport_run.run return value have both stdout and stderr? or am I understanding sth wrong? | 09:24 |
| openstack | Launchpad bug 1544537 in Solar "On error transport result should contain both stdout and stderr" [Medium,New] | 09:24 |
| openstackgerrit | Bogdan Dobrelya proposed openstack/solar: Add packer build for docker container https://review.openstack.org/281767 | 09:26 |
| openstackgerrit | Bogdan Dobrelya proposed openstack/solar: Add centos7 docker support https://review.openstack.org/284782 | 09:26 |
| pigmej | bogdando: I'm starting to think that our vagrantfile is too complicated already | 09:27 |
| pigmej | at least, it crashes a lot on fedora | 09:27 |
| pigmej | after your changes I get random errors ;/ | 09:27 |
| pigmej | `Name `solar_solar-dev1` of domain about to create is already taken.` like these | 09:27 |
| pigmej | I never ever saw them before | 09:27 |
| salmon_ | mkwiek: it's about this https://github.com/openstack/solar/blob/master/solar/core/transports/base.py#L81 | 09:29 |
| openstackgerrit | Bogdan Dobrelya proposed openstack/solar: Fix riak container nested docker mounts https://review.openstack.org/286501 | 09:30 |
| bogdando | pigmej, which changes? | 09:31 |
| pigmej | docker stuff | 09:31 |
| pigmej | I'm starting to think that packer + dockre is baaaaaaad idea ;/ | 09:32 |
| bogdando | pigmej, that is not the topic of the patch to w/a fixes | 09:32 |
| pigmej | bogdando: yeah I know :) | 09:32 |
| bogdando | to w/a bad ssh keys | 09:32 |
| pigmej | I know I dind't mean that patch exactly | 09:32 |
| bogdando | docker changes perhaps may be adapted better | 09:32 |
| pigmej | I mean, it break my system completely it seems | 09:32 |
| bogdando | or put to a separate vagrantfile | 09:32 |
| pigmej | after yesterday fuckups | 09:32 |
| bogdando | The packer + docker works for me well, on ubuntu trusty | 09:33 |
| bogdando | I can't say about fedora yet | 09:33 |
| pigmej | So for today I need to cleanup my system ;/ | 09:34 |
| mkwiek | salmon_: thanks | 09:34 |
| bogdando | pigmej, you could test it in ubuntu bv | 09:35 |
| bogdando | vm. And I can do the same for fedora vm | 09:35 |
| pigmej | bogdando: but if it can break the host system... it's bad | 09:35 |
| pigmej | I mean I don't get what went wrong yesterday about that logout/tty change | 09:36 |
| pigmej | it's just werid | 09:36 |
| bogdando | As I said, I have no details that was wrong and cannot reproduce. I don't have host fedora system | 09:36 |
| pigmej | I know ;/ | 09:36 |
| bogdando | I can only suggest to exchange test cases. Works for me at ubuntu - please confirm at ubuntu VM | 09:36 |
| pigmej | and well ubuntu image worked | 09:36 |
| bogdando | Doesn't work for you on fedore - I will try to repro at fedora vm | 09:36 |
| bogdando | makes sens? | 09:36 |
| pigmej | I mean fedora + centos7 => wtf | 09:36 |
| pigmej | fedora + ubutnu => ok | 09:37 |
| bogdando | let's just exchange test cases | 09:37 |
| pigmej | I can setup ubuntu VM and play with it | 09:37 |
| pigmej | :) | 09:37 |
| bogdando | please test ubuntu VM -> ubuntu/centos7 only. I will test fedora centos7 only | 09:37 |
| bogdando | as you already tested fedora->ubuntu | 09:38 |
| bogdando | if I can repro, we will just recommend to not use centos7 solar-dev containers at fedora | 09:38 |
| pigmej | bogdando: I can play with that stuff, but later, after we will close all stuf for F2S | 09:40 |
| bogdando | okay, although those docker containers can speed up our CI/dev a lot | 09:42 |
| pigmej | I know, but right now they are not working for me, so.. ;/ | 09:42 |
| pigmej | also they will not speed up CI at all right now, because we will need to adjust/remove/rewrite our current env, | 09:42 |
| pigmej | so they for sure can add speedup and they are more flexible, but they require a lot of time before we can fully benefit from them | 09:43 |
| bogdando | pigmej, do you think we could create another vagrantfile for docker? | 09:45 |
| pigmej | hmm, we could, but I'm not sure if it changes anything with docker itself :) | 09:46 |
| pigmej | for sure it would be more readable, and we could use `load` stuff to load shared parts | 09:46 |
| bogdando | that was to the topic to not get the Vagrantfile too scary | 09:46 |
| pigmej | yeah it's scary a bit and we stil have there only one iface for docker :D | 09:47 |
| bogdando | although I think the Vagrantfile may be complex, it is not for real deploys anyway :) only to play with | 09:47 |
| pigmej | Yeah | 09:48 |
| bogdando | I will add more interfaces support, but now it is pointless. We are not there yet | 09:48 |
| pigmej | To me it can be complex, but I'm starting to think that this vagrant + packer are not soo production ready if it comes to docker ;/ | 09:48 |
| bogdando | My point they never shall be for production | 09:48 |
| bogdando | only for dev/qa/ci | 09:49 |
| pigmej | surely, but for CI we also probably cannot use vagrant | 09:49 |
| bogdando | for production is must be only <package_manager> install solar | 09:49 |
| bogdando | and no prebuilt things more | 09:49 |
| pigmej | that's why salmon_ wrote our tests using fuel-devops stuff | 09:49 |
| bogdando | but I mean all of this vagrant, packer, docker, devops will never be related to solar in production | 09:50 |
| bogdando | we can do for our needs and dev velocity what we need | 09:50 |
| pigmej | yeah, BUT if we want to attract more peoples | 09:51 |
| pigmej | people* | 09:51 |
| bogdando | well packer, vagrant, docker will be the best | 09:51 |
| pigmej | we need good and easy dev env, and well, I'm starting to have impression that docker should be only "experimental" | 09:51 |
| bogdando | a laptop can run 10 docker containers of solar-dev nodes but not 10 vbox/libvirt vms | 09:52 |
| pigmej | well, looking at my fedora... I'm not sure if I would recommend docker to anyone | 09:52 |
| bogdando | yes, that was my intention, I noted the docker provider *is* experimental | 09:52 |
| pigmej | sure, but there are a lot of hacks for that "docker" stuff, and not to mention that my fedora (I didn't changed advanced system options) dislikes docker + vagrant a lot | 09:52 |
| bogdando | and looking to my ubuntu, I would :D | 09:52 |
| pigmej | I still don't get how the heck Centos7 docker can logout me from my machine.. | 09:53 |
| bogdando | yes, funny things | 09:53 |
| pigmej | I bet vagrant did something "wrong" | 09:53 |
| bogdando | I hope vagrant will fix missing docker networking and exec provisioners, that would make things more simple. | 09:54 |
| pigmej | and because docker have /sbin/init as run script.. | 09:54 |
| bogdando | And I will remove all of the "docker" hacks then. For now, it must be done | 09:54 |
| bogdando | anyway, I will use it for my own needs, as it works for me. But I hoped to leverage solar community, even if experimental and with temp w/a - it's damn fast | 09:56 |
| pigmej | I will check it for sure, because it's worth of having it but I need to focus on F2S for a bit | 09:56 |
| pigmej | yeah it's fast that's for sure | 09:56 |
| pigmej | :) | 09:56 |
| bogdando | and has nothing to the production use cases, as we all agree | 09:56 |
| bogdando | okay | 09:57 |
| pigmej | sure, but you know, I probabaly execute vagrant up like once a week | 09:57 |
| openstackgerrit | Merged openstack/solar: Fix riak container nested docker mounts https://review.openstack.org/286501 | 09:59 |
| pigmej | dshulyak: https://review.openstack.org/#/c/286665/ why it fails btw? | 10:07 |
| dshulyak | pigmej: tests are not adjusted | 10:07 |
| openstackgerrit | Bogdan Dobrelya proposed openstack/solar: Add packer build for docker container https://review.openstack.org/281767 | 10:16 |
| openstackgerrit | Bogdan Dobrelya proposed openstack/solar: Add centos7 docker support https://review.openstack.org/284782 | 10:16 |
| openstackgerrit | Maciej Kwiek proposed openstack/solar: Make SolarTransportResult output contain both outs https://review.openstack.org/287108 | 11:00 |
| openstackgerrit | Bogdan Dobrelya proposed openstack/solar: Add troubleshooting to the development guide https://review.openstack.org/287109 | 11:01 |
| openstackgerrit | Bogdan Dobrelya proposed openstack/solar: Add troubleshooting to the development guide https://review.openstack.org/287109 | 11:10 |
| openstackgerrit | Bogdan Dobrelya proposed openstack/solar: Add troubleshooting to the development guide https://review.openstack.org/287109 | 11:13 |
| openstackgerrit | Bogdan Dobrelya proposed openstack/solar: Add troubleshooting to the development guide https://review.openstack.org/287109 | 11:13 |
| openstackgerrit | Bogdan Dobrelya proposed openstack/solar: Add troubleshooting to the development guide https://review.openstack.org/287109 | 11:16 |
| dshulyak | salmon_: https://github.com/Mirantis/f2s/pull/17 | 11:28 |
| pigmej | salmon_: are there any special reqs for that mkdir pattern? | 11:28 |
| salmon_ | dshulyak: commented | 11:30 |
| salmon_ | pigmej: just create full path | 11:31 |
| dshulyak | ;) | 11:31 |
| pigmej | salmon_: k, | 11:32 |
| openstackgerrit | Merged openstack/solar: Add troubleshooting to the development guide https://review.openstack.org/287109 | 11:40 |
| pigmej | OSError: /usr/local/lib/python2.7/dist-packages/zmq/libzmq.so: undefined symbol: crypto_secretbox_open | 11:52 |
| *** dshulyak has quit IRC | 12:41 | |
| openstackgerrit | Jedrzej Nowak proposed openstack/solar: Create remote paths before syncing https://review.openstack.org/287170 | 12:51 |
| pigmej | salmon_: ^ | 12:51 |
| salmon_ | pigmej: I will test it now | 12:52 |
| salmon_ | pigmej: you can also think how to support this https://github.com/openstack/fuel-library/blob/master/deployment/puppet/osnailyfacter/modular/astute/tasks.yaml#L207-L208 :) | 12:53 |
| pigmej | in the same way in fact... | 12:55 |
| pigmej | how does fuel work with this ? | 12:55 |
| salmon_ | pigmej: https://github.com/openstack/fuel-astute/blob/master/mcagents/uploadfile.rb | 12:55 |
| salmon_ | it's run on slave node | 12:55 |
| pigmej | ah fuck | 12:56 |
| pigmej | mco :D | 12:56 |
| salmon_ | yes, this is mcollective agent | 12:57 |
| pigmej | so, our sync handler doesn't support this ;/ | 12:57 |
| salmon_ | yet :P | 12:57 |
| pigmej | it will be slow as hell | 12:57 |
| salmon_ | what? | 12:57 |
| salmon_ | I'm just talking about permissions | 12:58 |
| pigmej | well, any command will require full roundtrip | 12:58 |
| pigmej | run_transport(chmod...) | 12:58 |
| pigmej | run_transport(chmod...) | 12:58 |
| openstackgerrit | Merged openstack/solar: Fabric sync transport does not exit anymore https://review.openstack.org/285268 | 12:58 |
| salmon_ | pigmej: you can do it for all files at once | 12:59 |
| pigmej | or yet another sync_transport(soruce, dest) + sync_transport(file_with_all_permissions) + run_transport(execute permissions) | 12:59 |
| pigmej | solar agent could handle this perfectly | 13:00 |
| pigmej | BUT we're not going this way, so probably mco transport is not that bad idea right now | 13:00 |
| pigmej | because all we would need it to feed it with correct arguments | 13:00 |
| pigmej | because fuel does one request while syncing stuff, we will require more (or some "nice" hacks around rsync_path) :P | 13:01 |
| salmon_ | pigmej: for now we need something what works | 13:02 |
| pigmej | I can try but you know.... | 13:04 |
| pigmej | just let me know if this current patch fixes remote existence problem, then I will think how to add these permissins... | 13:05 |
| salmon_ | pigmej: ok, testing, It will take some time | 13:06 |
| pigmej | https://review.openstack.org/284293 we have Workflow+1 for our blocker in fuel :) | 13:16 |
| pigmej | nice ;) | 13:16 |
| salmon_ | and -2 from jenkins :( | 13:28 |
| pigmej | yaeh but it rather broke just becasue :P | 13:29 |
| openstackgerrit | Merged openstack/solar: Timestamp comparison in calculating time fixed https://review.openstack.org/286040 | 13:32 |
| *** dshulyak has joined #openstack-solar | 13:32 | |
| pigmej | salmon_: how is testing going ? | 13:40 |
| salmon_ | it's going :) | 13:41 |
| dshulyak | guys, do oyu have fix for rsync? | 13:46 |
| dshulyak | i was disconnected for a while | 13:46 |
| dshulyak | salmon_: pigmej ^ ? | 13:46 |
| openstackgerrit | Maciej Kwiek proposed openstack/solar: Change PENDING_RETRY to ERROR_RETRY in soft_stop https://review.openstack.org/287044 | 13:48 |
| dshulyak | ok, i will try this one https://review.openstack.org/#/c/287170/ | 13:48 |
| pigmej | dshulyak: we do | 13:49 |
| pigmej | exactly this one :) | 13:49 |
| pigmej | then salmon_ said me about: https://github.com/openstack/fuel-library/blob/master/deployment/puppet/osnailyfacter/modular/astute/tasks.yaml#L207-L208 | 13:49 |
| dshulyak | yeah, there is timeouts and retries almost for every task, maybe it will be appropriate to handle them on transports layer | 13:51 |
| pigmej | dshulyak: and also: https://github.com/openstack/fuel-library/blob/master/deployment/puppet/osnailyfacter/modular/astute/tasks.yaml#L194-L195 | 13:52 |
| pigmej | dshulyak: because this file just changed :D | 13:53 |
| pigmej | and the 207-208 was pointing to these "permissions" | 13:53 |
| dshulyak | :) | 13:53 |
| pigmej | btw how do you guys plan to pass the permissions stuff ? | 14:08 |
| pigmej | let's say that it that {src, dst} dict ? | 14:08 |
| pigmej | like {src, dst, permissin} ? | 14:08 |
| salmon_ | pigmej: permision is for whole list | 14:10 |
| pigmej | always? | 14:10 |
| salmon_ | it looks like | 14:10 |
| dshulyak | yes, atleast for existing tasks | 14:11 |
| dshulyak | btw my primary-controller still runs, cool | 14:11 |
| pigmej | werid | 14:11 |
| dshulyak | salmon_: have u tried with compute ? | 14:11 |
| pigmej | I mean, I don't know then how we want to support this ;/ | 14:11 |
| salmon_ | dshulyak: nope I started from th e begining and provision failed on one node. Trying to fix | 14:13 |
| pigmej | salmon_: can you merge this rsync stuff? | 14:16 |
| pigmej | then I can start working on permissions | 14:16 |
| pigmej | we should somehow create tests for handlers... | 14:16 |
| salmon_ | pigmej: when it's tested :) | 14:17 |
| openstackgerrit | Bogdan Dobrelya proposed openstack/solar: W/a virtualbox hardcoded paths for solar examples https://review.openstack.org/286730 | 14:17 |
| mkwiek | dshulyak: what's up with ctxt? | 14:21 |
| mkwiek | is it some kind of interface that we want to implement with Scheduler objects? | 14:21 |
| pigmej | salmon_: dshulyak already did +2 :P | 14:21 |
| dshulyak | mkwiek: every public method of worker should be able to accept ctxt | 14:21 |
| mkwiek | dshulyak: yeah, you wrote this in comment, but can you provide some more info? I am not saying you are wrong, I am saying that I don't understand why this is the case :_) | 14:23 |
| mkwiek | *:) | 14:23 |
| dshulyak | mkwiek: this is just how i implemented subscribers https://github.com/openstack/solar/blob/master/solar/orchestration/workers/subscription.py#L55-L83 for workers | 14:25 |
| dshulyak | ctxt may have info that is not required by task, but might be needed for other purposes | 14:25 |
| mkwiek | dshulyak: thanks, that's the info I needed, will revert these lines :) | 14:25 |
| dshulyak | primary-database failed | 14:27 |
| openstackgerrit | Maciej Kwiek proposed openstack/solar: Change PENDING_RETRY to ERROR_RETRY in soft_stop https://review.openstack.org/287044 | 14:28 |
| salmon_ | mkwiek: https://bpaste.net/show/67bbbbca22ba | 14:32 |
| salmon_ | is it somehow connected with your changes? | 14:32 |
| pigmej | I have seen it before afair | 14:34 |
| mkwiek | salmon_: not sure, it didn't happen to me when I was trying out my changes | 14:35 |
| pigmej | dshulyak: what we do about that versioning thingy? | 14:35 |
| mkwiek | does it happen immediately after calling cli? | 14:35 |
| salmon_ | pigmej: it's working | 14:37 |
| pigmej | salmon_: merge then :P | 14:38 |
| dshulyak | pigmej: lets do this tommorow, i am trying to understand why galera fails and i need to fix tests for my other patch | 14:39 |
| pigmej | dshulyak: ok | 14:40 |
| pigmej | I will add workflow -1 to it | 14:40 |
| pigmej | "just in case" :) | 14:40 |
| dshulyak | bogdando: are u around? i have this error in puppet log - G2016-03-02 14:43:57 +0000 Haproxy_backend_status[mysql](provider=http) (debug): Got error while checking backend: No route to host - connect(2) | 14:44 |
| dshulyak | bogdando: primary-database task | 14:45 |
| salmon_ | dshulyak: I'm trying now controller+compute | 14:46 |
| bogdando | dshulyak, those primary- ones are better to be consulted with authors, Matthew perhaps | 14:46 |
| bogdando | they split roles, I'm not aware off related issues | 14:47 |
| bogdando | but I can try to debug your env... | 14:47 |
| pigmej | salmon_: dshulyak if permission says dir_permissions it's chmod {permission} dir or (I don't think so) chmod -R {permission} dir ? | 14:50 |
| pigmej | I also see that some tasks have `'cwd': /` we don't support it too | 14:51 |
| dshulyak | pigmej: chmod dir, and then chmod for each file - https://github.com/openstack/fuel-astute/blob/master/mcagents/uploadfile.rb#L37-L44 | 14:51 |
| dshulyak | pigmej: i remember that it was important for 1st verions of plugins, we will need to recheck it later | 14:52 |
| pigmej | there are a lot of custom parameters that we don't support ;/ | 14:52 |
| dshulyak | retries + interval / timeout / cwd ? | 14:53 |
| pigmej | chown, chgrp | 14:54 |
| pigmej | dshulyak: I can implement them though, just time needed, Next time please communicate these things asap :) | 15:01 |
| pigmej | I mean not only dshulyak :) salmon_ also :P | 15:02 |
| salmon_ | pigmej: I don't think that it's important for now :P | 15:03 |
| salmon_ | w sumie na jobvite nie potwierdzil | 15:04 |
| pigmej | salmon_: yeah I know | 15:04 |
| pigmej | next time switch window :) | 15:04 |
| salmon_ | :D | 15:04 |
| pigmej | salmon_: what is not important ? | 15:07 |
| pigmej | permissions for sure are, because apache may not start with worng permissions to ssl things | 15:08 |
| salmon_ | retries + interval / timeout / cwd | 15:08 |
| pigmej | first 3 probably | 15:08 |
| pigmej | but 'cwd' I don't know, we always run from $HOME in fact | 15:08 |
| pigmej | so it may be problematic, | 15:09 |
| salmon_ | we will see | 15:09 |
| salmon_ | I didn't see errors from it yet, but I'm looking at some failures :) | 15:10 |
| pigmej | cwd is the easiest from these all | 15:12 |
| pigmej | but cwd is for run transport | 15:12 |
| pigmej | and I think we support it there | 15:13 |
| pigmej | ;d | 15:13 |
| pigmej | yup we do | 15:13 |
| pigmej | so cwd is not an issue ;) | 15:13 |
| openstackgerrit | Aleksey Zvyagintsev proposed openstack/solar-resources: Add spec for rpmbuild solar-resources.spec https://review.openstack.org/287256 | 15:16 |
| mkwiek | pigmej: can you elaborate a bit on this bug: https://bugs.launchpad.net/solar/+bug/1549342 ? I am not sure how to reproduce it | 15:20 |
| openstack | Launchpad bug 1549342 in Solar "Fix python interpreter errors on interpreter shutdown" [Undecided,New] - Assigned to Jedrzej Nowak (jnowak) | 15:20 |
| pigmej | it's tricky beast :D | 15:21 |
| pigmej | you should enable pg backend probably | 15:22 |
| pigmej | shit it doesn't want to crash again on PG | 15:22 |
| pigmej | mkwiek | 15:22 |
| pigmej | the thing is that if you have program that uses gevent and you patch gevent too late, you *may* have some weird exceptions at the end | 15:23 |
| mkwiek | ok, I think I will leave this one, but maybe I will mark it as incomplete on LP until we find a proper way to reproduce it? | 15:24 |
| mkwiek | pigmej: also, when does this one happen: https://bugs.launchpad.net/solar/+bug/1551180 ? | 15:27 |
| openstack | Launchpad bug 1551180 in Solar "ImportError: cannot import name sugar" [Undecided,New] | 15:27 |
| mkwiek | there is no info in the bug :( | 15:27 |
| pigmej | well for me it happens everytime when I start new machine... | 15:28 |
| pigmej | it's some error with zmq stuff | 15:28 |
| pigmej | I don't know wtf, but it looks like you need to reinstall zmq | 15:29 |
| pigmej | (uninstall in fact first) then install it again | 15:29 |
| mkwiek | o_O | 15:29 |
| pigmej | and it happens on any solar command | 15:29 |
| pigmej | basically if you will try to import zmq => error | 15:29 |
| pigmej | mkwiek: https://blueprints.launchpad.net/solar/+spec/rework-bat-transport-stevedore you could try this one | 15:30 |
| pigmej | but it's not simple | 15:30 |
| pigmej | it's tricky, and it's not a bug, but fixing "bad" idea after porting to stevedore | 15:31 |
| pigmej | or obvioulsy, you can play with CLI, there is probably milion bugs there | 15:31 |
| pigmej | but not reported (yet) | 15:31 |
| mkwiek | ok, thanks for this, I will check this out | 15:32 |
| pigmej | and because you started playing with solar you may have some fresh look on CLI | 15:34 |
| pigmej | CLI is critical for us because it will be only one UX that we will support in 9.0 | 15:36 |
| pigmej | also https://blueprints.launchpad.net/solar/+spec/repository-module-atomic-like-create this one is maybe even more required than that BAT stevedore fixes | 15:36 |
| salmon_ | dshulyak: I think we are missing between nodes events | 15:38 |
| mkwiek | pigmej: ok, I am on the one with repository creation | 15:38 |
| dshulyak | salmon_: can u check it with any task that have cross-depends ? | 15:39 |
| salmon_ | dshulyak: I will, for now checking what's working | 15:40 |
| pigmej | mkwiek: when you will have some PoC / idea please share it with me, k? | 15:40 |
| mkwiek | sure | 15:40 |
| dshulyak | pigmej: what is the easiest way to add input for resource? | 15:45 |
| dshulyak | i need to create location_id somehow | 15:45 |
| pigmej | dshulyak: solar input add ? | 15:45 |
| dshulyak | in python | 15:45 |
| pigmej | resource.add_input | 15:45 |
| salmon_ | +1 | 15:45 |
| pigmej | input_add | 15:46 |
| pigmej | dshulyak: or if you don't have resource but dbobj | 15:46 |
| pigmej | something like dbobj.inputs.add_new(name, value, schema=None) | 15:46 |
| pigmej | schema=None => will be guessed | 15:46 |
| salmon_ | dshulyak: I'm using it here https://github.com/Mirantis/f2s/blob/master/f2s/fsclient.py#L191 | 15:46 |
| pigmej | but also | 15:46 |
| pigmej | location_id *should* be created automaticaly | 15:46 |
| pigmej | IF you use Resource.class | 15:47 |
| pigmej | there is Resource.auto_extend_inputs | 15:47 |
| pigmej | which adds that stuff | 15:47 |
| dshulyak | Resource.class reads from filesystem afair | 15:47 |
| dshulyak | i will add it using schema | 15:47 |
| dshulyak | i mean dbobj.inputs.add_new | 15:47 |
| pigmej | you can :) | 15:47 |
| pigmej | well, resource.from_spec uses filessytem | 15:47 |
| pigmej | but resource.from_db is not :) | 15:48 |
| pigmej | :) | 15:49 |
| salmon_ | compute registered, nice :) | 15:49 |
| pigmej | cool! | 15:49 |
| pigmej | btw why the f** that https://github.com/openstack/solar-resources/blob/master/resources/sources/1.0.0/meta.yaml is named SOURCE ? | 15:50 |
| salmon_ | hmm https://readthedocs.org/projects/solar/builds/ | 15:51 |
| salmon_ | all failes | 15:51 |
| pigmej | nice :D | 15:52 |
| salmon_ | nope | 15:52 |
| pigmej | who broke it and why it was me? | 15:52 |
| pigmej | but it's weird | 15:53 |
| pigmej | https://github.com/openstack/solar/blob/master/solar/version.py | 15:53 |
| salmon_ | ImportError: No module named solar.version | 15:53 |
| pigmej | salmon_: that's weird | 15:56 |
| pigmej | because it exists | 15:56 |
| salmon_ | pigmej: yup, wtf moment ;) | 15:57 |
| pigmej | and you can import that version from solar/ dir | 15:57 |
| pigmej | but you cannot solar.version from outside.. | 15:57 |
| salmon_ | why not? | 15:58 |
| pigmej | idk | 15:58 |
| pigmej | let's ignore it | 15:58 |
| pigmej | I will submit other patch that fixes it | 15:58 |
| pigmej | dshulyak: Patch Set 18: Verified+2 :D | 16:02 |
| pigmej | you got your long awaited merge :) | 16:03 |
| pigmej | so we don't need FFE :D | 16:03 |
| *** ikalnitsky_ has joined #openstack-solar | 16:13 | |
| dshulyak | pigmej: it was long journey | 16:15 |
| pigmej | yeah like in fuel :D | 16:15 |
| dshulyak | long journey from home to store | 16:15 |
| pigmej | dshulyak: remember to cancel FFE :) | 16:15 |
| *** ikalnitsky has quit IRC | 16:15 | |
| pigmej | dshulyak: good job with that patch :) | 16:15 |
| *** ikalnitsky_ is now known as ikalnitsky | 16:16 | |
| dshulyak | pigmej: yes, i will send a note | 16:17 |
| salmon_ | dshulyak: and add some gif http://giphy.com/gifs/no-floyd-mayweather-manny-pacquiao-3oEduZr3O04BtJNWSs | 16:18 |
| pigmej | :D | 16:20 |
| salmon_ | dshulyak: how can I test it https://review.openstack.org/#/c/286665/ ? | 16:22 |
| dshulyak | salmon_: u can test it with f2s :) for example without this patch if you commit anchors before main deployment (fuel alloc 1 null) | 16:23 |
| dshulyak | and then u will try to fuel alloc 1 1 - u will see that graph is disconnected | 16:23 |
| salmon_ | hmm | 16:23 |
| salmon_ | I think I did | 16:24 |
| salmon_ | *I did it | 16:24 |
| salmon_ | dshulyak: ok, will check | 16:24 |
| salmon_ | after debugging swift | 16:24 |
| dshulyak | for example u wont have deploy_end anchor, and all tasks that needs to be run after this anchor - can be executed before any task | 16:25 |
| salmon_ | hmm, I had it :) | 16:25 |
| *** salmon_1 has joined #openstack-solar | 16:29 | |
| *** salmon_ has quit IRC | 16:30 | |
| *** dshulyak has quit IRC | 16:36 | |
| *** salmon_1 is now known as salmon_ | 16:45 | |
| *** dshulyak has joined #openstack-solar | 17:15 | |
| *** dshulyak has quit IRC | 17:20 | |
| *** dshulyak has joined #openstack-solar | 17:40 | |
| pigmej | hmm | 18:07 |
| pigmej | guys I have question to you, `self.transport_sync.copy(resource, item['src'], item['dst'])` | 18:08 |
| pigmej | why there is no 'sudo' ? | 18:08 |
| pigmej | you're using user that has access to these files "always" ? | 18:09 |
| pigmej | salmon_: | 18:09 |
| pigmej | dshulyak: | 18:09 |
| salmon_ | pigmej: hmm, we are using root :) | 18:09 |
| dshulyak | btw | 18:09 |
| dshulyak | there is non-root feature, and ssh with root will be disabled | 18:10 |
| dshulyak | but | 18:10 |
| dshulyak | there will be service user with sudo and ssh key | 18:10 |
| pigmej | ok, because my next question was how the heck are we going to solve it | 18:10 |
| pigmej | because I wrote that chown/chgrp stuff | 18:10 |
| pigmej | and obviously user cannot chown | 18:11 |
| pigmej | except when user == root | 18:11 |
| pigmej | obviously | 18:11 |
| dshulyak | i didnt get it, we will be able to use sudo without password | 18:12 |
| pigmej | yeah but that's the point | 18:12 |
| pigmej | if you login as root => you don't need to use sudo in a *lot* of places | 18:12 |
| pigmej | when you're not a root => you need to | 18:12 |
| pigmej | so my question is, how the heck are we going to solve this :) | 18:12 |
| dshulyak | can we add sudo: yes to our ssh transport ? | 18:13 |
| dshulyak | or become :D | 18:13 |
| pigmej | yeah I wanted to do so, BUT, it's not that trivial change | 18:13 |
| pigmej | because all "sudo" ifs needs to be adjusted etc | 18:13 |
| pigmej | so I would prefer to have this at naive_sync logic | 18:14 |
| pigmej | at least for now | 18:14 |
| pigmej | https://bpaste.net/show/4a3ae32c4da9 | 18:14 |
| pigmej | then full blown example could look like | 18:14 |
| pigmej | OR | 18:16 |
| dshulyak | without sudo: yes in ssh transport - how we are going to run puppet/bash/etc ? | 18:16 |
| pigmej | https://bpaste.net/show/f582d90231ed | 18:16 |
| pigmej | dshulyak: logic is in handlers | 18:16 |
| pigmej | handler is adding sudo when needed | 18:16 |
| pigmej | so in fact puppet sets always to True, while ansible uses "become" stuff | 18:17 |
| pigmej | https://github.com/openstack/solar/blob/master/solar/core/handlers/puppet.py#L57 etc | 18:17 |
| pigmej | this logic is not on transport level, from transport we assume that you have what handler needs, and handler determines what to do, it's cool approach, because we don't need to care about "sudo" at transport level, so handler executes everything with correct stuff | 18:18 |
| pigmej | and because we use usually (I'm not sure if in all palces) transport.run(...use_sudo=True) we can swap the "sudo" implementation on transport level. | 18:19 |
| dshulyak | yes, it makes sense | 18:19 |
| pigmej | BUT for naive_sync, we lack information when to sudo, when not | 18:19 |
| pigmej | so I have to add it somewhere | 18:19 |
| pigmej | https://bpaste.net/show/5e1f41d46263 | 18:19 |
| pigmej | this is 3rd version, I think it's the best | 18:19 |
| pigmej | BUT there is a problem with owner stuff, which always requires sudo | 18:20 |
| dshulyak | to me it makes more sense to store this in inputs, because if u can update source - then sudo input may change as well | 18:20 |
| openstackgerrit | Merged openstack/solar: Make SolarTransportResult output contain both outs https://review.openstack.org/287108 | 18:21 |
| dshulyak | so 1 or 3 i guess | 18:21 |
| *** angdraug has joined #openstack-solar | 18:21 | |
| pigmej | k, | 18:22 |
| pigmej | what if sudo=False and you set chown ? | 18:22 |
| pigmej | `chown: changing ownership of ‘a’: Operation not permitted` ? | 18:23 |
| pigmej | :) | 18:23 |
| dshulyak | it looks like either all of additonal options should be per file or none | 18:23 |
| pigmej | well, there is also one other case | 18:24 |
| pigmej | always execute chgrp/chown with sudo | 18:24 |
| dshulyak | :) | 18:25 |
| dshulyak | what if u dont have sudo, and u will try to run those commands ? | 18:27 |
| dshulyak | i think chown/chgrp should be executed if sudo: yes or ssh user is root | 18:27 |
| pigmej | ok, so I will do this way and I will put note there, | 18:29 |
| pigmej | later we may change implementation | 18:30 |
| pigmej | :) | 18:30 |
| pigmej | well if yuo don't have sudo there is still a chance that you can change group | 18:30 |
| pigmej | I will put sudo for now at the same level as group things | 18:30 |
| pigmej | to keep it consistent, later we can create more granular stuff | 18:31 |
| pigmej | shit it makes completely no sense... | 18:34 |
| pigmej | https://bpaste.net/show/720ea5b766f1 | 18:50 |
| pigmej | dshulyak: salmon_ | 18:50 |
| pigmej | https://bpaste.net/show/3612abacee53 | 18:51 |
| pigmej | :) | 18:51 |
| salmon_ | cool :) | 18:53 |
| pigmej | you haven't yet seen the code :D :D | 18:53 |
| salmon_ | https://bpaste.net/show/b0148c02a98f :( | 18:58 |
| pigmej | salmon_: yup | 18:59 |
| pigmej | I just reported this | 18:59 |
| pigmej | but I think | 18:59 |
| pigmej | https://review.openstack.org/#/c/286571/ this fixes it | 18:59 |
| salmon_ | but my graph is not empty :( | 19:00 |
| pigmej | it's exactly this bug | 19:01 |
| pigmej | just different "reason" | 19:01 |
| pigmej | it will be fixed by that one | 19:01 |
| pigmej | :) | 19:01 |
| openstackgerrit | Jedrzej Nowak proposed openstack/solar: Support permissions, owner and group in naive_sync https://review.openstack.org/287390 | 19:11 |
| openstackgerrit | Jedrzej Nowak proposed openstack/solar-resources: Use new atributes in sources resource https://review.openstack.org/287394 | 19:14 |
| pigmej | salmon_: dshulyak first one allows you to use attributes from the second one :) | 19:14 |
| *** dshulyak has quit IRC | 21:06 | |
| *** angdraug has quit IRC | 21:45 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!