*** sylwesterB has joined #openstack-bareon | 07:41 | |
*** sylwesterB has quit IRC | 07:44 | |
*** sylwesterB has joined #openstack-bareon | 07:44 | |
*** sylwesterB_ has joined #openstack-bareon | 09:33 | |
*** sylwesterB has quit IRC | 09:34 | |
*** sylwesterB_ is now known as sylwesterB | 09:34 | |
evgenyl | sylwesterB: hi, could you please rebase your patch https://review.openstack.org/#/c/260453/ since you don't need changes in requirements.txt, and there is typo-comment from Bulat :) | 09:45 |
---|---|---|
sylwesterB | yeah, ok | 09:55 |
sylwesterB | evgenyl: Does the customized order in pipelines is important? | 10:50 |
evgenyl | sylwesterB: it's a good question. | 10:51 |
evgenyl | sylwesterB: ok, for now lets sort by name, in order to have deterministic output, but as a possible future improvment add info that it will be possible to implement custom order using integer field, when there are such cases. | 10:52 |
evgenyl | sylwesterB: I'm not sure if it will be needed, these are not plugins, so we shouldn't override some data, just add new. | 10:53 |
sylwesterB | evgenyl: I'm also wondering i a pipeline should be a class with one 'main' method like "run" and it will call all other methods | 10:57 |
sylwesterB | and if we want to change the order we should override "run" in subclass | 10:57 |
sylwesterB | OR | 10:57 |
sylwesterB | define pipeline like in i.e. python social auth | 10:57 |
sylwesterB | as a list of functions (string), which are loader dynamicaly (by importlib but we would use stevedore) and run one by one in some for loop | 10:58 |
evgenyl | sylwesterB: hm, I think it's better to have a method, and on level higher go get the data through pipeline which is build in name-based-order. | 10:59 |
evgenyl | sylwesterB: didn't get your suggestion, could you please write some pseudo-code to see how it would looks like? | 10:59 |
sylwesterB | ok | 11:00 |
sylwesterB | evgenyl: http://paste.openstack.org/show/484982/ | 11:07 |
sylwesterB | this is how it looks in PSA | 11:08 |
sylwesterB | more or less | 11:08 |
agordeev | hi guys. meanwhile decoupling do_actions from manager into separate files, i found new type of objects which we might need to introduce into bareon. | 11:11 |
agordeev | at first i would like to show an example of action which can be separated as is. https://github.com/openstack/bareon/blob/master/bareon/manager.py#L333-L390 | 11:12 |
agordeev | now take a look at the other action which can't be smoothly separated https://github.com/openstack/bareon/blob/master/bareon/manager.py#L654-L769 | 11:13 |
agordeev | two methods from manager prevents that: mount_target() and umount_target() https://github.com/openstack/bareon/blob/master/bareon/manager.py#L447-L488 | 11:14 |
agordeev | what do you think about adding new type of objects named as 'helpers'? those mentioned method are the first candidates for helpers. | 11:16 |
evgenyl | sylwesterB: http://paste.openstack.org/show/484985/ I'm not sure if we should provide serializer classes are organized as an interface, we may change them easily in the future, on the other hand we have output of serializers, which is already versioned. | 11:19 |
agordeev | so, the helpers could be reused in multiple actions like utils. But it's a piece of code which needs access to data objects from driver. That will effectively distinct a helper from util. As utils should never use data object directly. | 11:20 |
evgenyl | agordeev: makes sense, but describe those definitions in doc-strings. | 11:21 |
agordeev | the amount of code and especially the amount of action which a helper will perform aren't enough to become a pure single action. | 11:22 |
evgenyl | agordeev: so utils is a layer which is closer to OS, higher we have helpers, is it correct? | 11:22 |
sylwesterB | btw mount_target and unmount_target looks like ideal candidate to be convert into context manager ;) | 11:23 |
agordeev | evgenyl: yes, it's absolutely correct | 11:23 |
sylwesterB | evgenyl: So you suggest to not interrupt current serialization process, and just add some data after it ends? | 11:25 |
evgenyl | sylwesterB: yep, lets do it stupidly simple, if we get more complicated cases - we will improve it, according to those cases, complication will cause a lot of questions :) | 11:26 |
sylwesterB | That's a shame. It makes this task less exciting :( | 11:28 |
sylwesterB | but OK ;) | 11:28 |
sylwesterB | I will prepare some PoC | 11:28 |
sylwesterB | When you come here? | 11:29 |
evgenyl | sylwesterB: :) | 11:31 |
evgenyl | sylwesterB: in 3 hours I think I'll be in Poznan office, sitting in Warsaw airport waiting for my flight. | 11:32 |
sylwesterB | on which one? | 11:32 |
sylwesterB | Modlin or Chopin? | 11:33 |
evgenyl | sylwesterB: oh, there are several airports! :) I'm in Chopin. | 11:34 |
sylwesterB | hmm | 11:40 |
sylwesterB | so the same as the train ;) | 11:41 |
*** openstackgerrit has quit IRC | 17:17 | |
*** openstackgerrit has joined #openstack-bareon | 17:17 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!