You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@ariatosca.apache.org by Digest <sl...@digest.ai> on 2017/05/18 05:01:19 UTC
Digest for ASF's Slack
Here’s your digest for today!
#ariatosca
undefined: <@U4RAXBZ1Q> I tried using <ma...@apache.org> but seems that the email doesn't get anywhere
Does the alias exist?
undefined: Just saw it in moderation give me a few mins
undefined: or we have to manually create it
oh cool
undefined: Busy watching <@U53P75YTD> pace around the speaker area
undefined: Is there a video stream or somthing ?
undefined: If I had to guess, audio
This is one of the few rooms without video, but I haven’t been to many of the big data sessions.
The wifi’s dropping out.
I just moderated the message
undefined: mehhhhh apacheCon has no video recording
make do with low quality audio
is there a decent crowd for <@U53P75YTD> talk ?
undefined: There was video in my talk
maybe its just a couple of rooms
undefined: they don’t have video recording in apachecons
except for keynotes
undefined: Unfortunately very small attendance. Must have been competing with a more popular topic. Also didn't help that time slot was changed twice
undefined: ^^^^ a big reason i hate apachecons
they push u out so that some big honcho’s last min talk has a higher priority
sorry to hear that <@U53P75YTD>
undefined: It's ok, even a little visibility is good
undefined: the problem is there r only about 300 attendees at any ApacheCon
and they have about 7 parallel talks happening
assuming that 70 out of 300 attendees r only there to enjoy Miami Beach
and 30 of the remaining are organizers
50 of the remaining are just there to sit in hallways
u have 150 attendees who are curious about sessions
undefined: Yes, also the context was a bit awkward - it's not strictly speaking a big data project
undefined: and break that across 7 parallel sessions
u can expect less than 20 people to show up
and assuming that there is a Spark or TensorFlow talk happening at the same time as urs
40 people go to that talk
undefined: If you knew all that why did you want me to come? :) J/k, exposure has to start somewhere
undefined: and so u end up with less than 10-15 people
yeah i knew all this
but this podling needed some little bit of exposure
hence i pushed u guys hard
undefined: [UPDATE] removed runtime properties entirely, added a more comprehensive support for attributes
undefined: <@U546PEP9S> this looks like the right move. how challenging will it be to convert plugins to using attributes? how are we going to solve the problem that attributes need to be pre-declared at the node types?
undefined: Which plugins are we talking about? The current solution enables to treat the attributes as if they were a dictionary (without .value) when accessing them from the operation ctx , this could be easily extended to workflow ctx. About the pre declaration, this is something we need to discuss, the obvious way is to place some "workspace" attribute which is not very TOSCA
Oh, and maybe more importantly, setting an value to an attribute is easy: `node.attributes['key'] = 'value'`
Behind the scenes this would create a parameter
Note that `node.attributes['key']['inner_key'] = 'value'` would create a single parameter which holds a dictionary `{'inner_key': 'value'}`
undefined: I didn't have any plugin in mind, it's just what i heard from Ran... Your solution is correct, as long as there's still a way for users to access/change the type name and description...
undefined: I was giving a try to <#C53MGE0DA|ariatosca> and I got:
```
$ aria --version
ARIA CLI 0.1.0
```
and the help
```
$ aria --help
(...)
Commands:
executions Handle workflow executions
logs Show logs from workflow executions
node-templates Handle a service template's node templates
nodes Handle a service's nodes
plugins Handle plugins
reset Reset ARIA's working directory
service-templates Handle service templates on the manager
services Handle services
workflows Handle service workflows
```
so, when I follow the docs:
```
$ aria parse blueprints/tosca/node-cellar/node-cellar.yaml
Usage: aria [OPTIONS] COMMAND [ARGS]...
Error: No such command "parse".
Did you mean one of these?
reset
```
any ideas?
installed using: `pip install git+<http://git-wip-us.apache.org/repos/asf/incubator-ariatosca.git>`
undefined: <@U53P75YTD> it is still possible, though I'm not sure it should be :wink:
undefined: Well, users should be able to create new attributes or modify them dynamically. Or maybe set a value according to its type. Or print out the description in a log file.
So how do you access the "raw" parameter model of an attribute?
undefined: From within an operation you can still do it, although it is a bit trickier...
undefined: Maybe it should be attribute_values in addition to attributes?
undefined: Its something like that
It's just _attributes instead of attributes
undefined: <@U53P75YTD>: any ideas _how to fix getting started with ariatosca_ based on our earlier discussion?
undefined: Ok, i mostly want to make sure the models are consistent
undefined: No changes were made to the models...
undefined: <@U5E39H8HE> git clone, pip install, try the examples for start :)
undefined: Apart from removing runtime properties that is
undefined: <@U546PEP9S> except for this attr. I would be more comfortable if the original attributes remained, and this new one would be attribute_values.
undefined: The attributes suggaring is at ctx level
undefined: Ohhh, ok. So not at modeling package?
undefined: Nope
undefined: Guys, Alex from Apifocal is very interested in learning about ARIA and seeing how he can help. Please be nice to him!
undefined: Hey <@U5E39H8HE>
undefined: [UPDATE] As part of the effort to convert the `Parameter` relationships from many-to-many to polymorphic one-to-many, I'm considering the creation of separate tables for inputs, output, properties and attributes. Did preliminary discussions with maxim and ran today, but before deciding on a model change, we'll discuss it together tomorrow evening (Israel time) to better think of the implications.
undefined: <@U53FAA5HP> why not just a foreign key per relation? The big advantage of a single table is that it would simplify the coerce_values phase: a single run on all rows
undefined: There are additional considerations. I'll explain more tomorrow =)
undefined: Ok. :) Boarding flight now, smell ya later
undefined: Hi <@U5E39H8HE> Welcome :slightly_smiling_face:
the Parse command is now replaced with
aria service_templates store [uri] -n [name]
aria service_templates show [name]
Parse existed before we added a storage layer to ARIA
this issue was reported on Jira(<https://issues.apache.org/jira/browse/ARIA-212>) and we have another Jira for updating the README file (<https://issues.apache.org/jira/browse/ARIA-166>)
Generally speaking, the flow using ARIA would be
1. to store a service template from the TOSCA YAML
# aria service-templates store helloworld.yaml hello
2. Create a service
# aria services create hello -t hello
3. Execute a workflow (the built in install workflow is a graph based wokflow, that traverses thru lifecycle interfaces of all nodes and relationships declated in the TOSCA template)
# aria executions start install -s hello
<@U5E39H8HE> you can find some into information here
<https://docs.google.com/presentation/d/1acxm-_fRyGbk4-o7EHHhtBdwIPsKHoPnMcqbwyh-grs/edit#slide=id.g18dc0515b2_0_0>
We are doing an online webinar tomorrow <http://cloudify.co/webinars/aria-model-driven-orchestration-webinar.html#register>
You are most welcome to join :slightly_smiling_face:
Oh, I just noticed I haven't switched to my account own (this is <@U53NR8XFF> ) :flushed:
undefined: <@U5E39H8HE> I hope this is helpful
undefined: Thanks <@U53NR8XFF> ... I'll take a look
Maybe even address #ARIA-166
undefined: That would be awesome! You can take a look at the contribution guild <https://cwiki.apache.org/confluence/display/ARIATOSCA/Contributing+to+ARIA>
#random
#general