Can we assign task to a group of workers ?

[thibaultduponchelle]

More on sparrow side, but can we assign a project to a group of workers, let's say "big nodes"... And some workers defined elsewhere as being "big nodes" can handle the build ?

[melezhik]

so, you want dynamically resolve a worker name/type ( rather then statically through a sparky.yaml ), right? If so, when do and how do you want to do this? if you please give me more context?

if this something could be achieved by FTP mechanism? A "supplemental" web application which handle a request will make a decision (when write up a tigger file) about what worker a build has to be run on ...

[thibaultduponchelle]

Not dynamically, more dumbly like a round robin.

But the question is more, is there a feature in Sparrow6 to run a job:

• On a all host defined in a group
• On one host of the group

[melezhik]

But the question is more, is there a feature in Sparrow6 to run a job:
On a all host defined in a group

it's possible to run sparrowdo scenario on group of hosts, by using a Sparky backend using hosts file syntax.

It's even possible to have tags semantic allowing sub grouping within groups and assigning named attributes to hosts ( ala chef nodes attributes ) ...

We gradually uncover all the Sparky/Sparrowdo/Sparrow treasures through these tickets, ha-ha 😄

[thibaultduponchelle]

Ahah yes !

This is great

[melezhik]

it's of course not a pure Sparky solution, but still does more or less groups of hosts asynchronous deployment. It's quite simple right now, not even a round robin, just a raku rand function, that is it - https://github.com/melezhik/sparrowdo/blob/master/bin/sparrowdo#L281 , anyway it does it's job ...

[melezhik]

Now with Sparky Job API, cluster mode you can run certain tasks on certain workers, please see https://github.com/melezhik/sparky#cluster-jobs

[thibaultduponchelle]

Looks good to me 😃