This message was deleted.

for example: do a mylvmsnapshot on many hosts in parallel. On those that were successful, run a mariadb upgrade plan, on the rest, return an error message, so that I may investigate and rerun

I don’t have anything under my hands now but I did similar things on my ex-job.

Yeah, once you have a

ResultSet

with all the targets you can use

.error_set

and

.ok_set

to filter out the ones that failed vs. succeeded respectively

thanks, that is pretty much what we settled on. I am always looking for ways to improve it though. for instance, if the plan includes a reboot, ALL hosts have to come back before ANY hosts can do the next thing. Is parallelize still considered a beta feature?

There are some examples: https://github.com/puppetlabs/puppetlabs-peadm/blob/main/plans/subplans/install.pp

i like this idea of a sub-plan way too many examples here where we havea 'task' that does too many things

as far as I see both

parallelize

and

background

are GA now

Parallelize is more or less here to stay, but note if you are planning on using this with PE those functions won't work there as of right now

you can do

parallelize($targets)

then you’ll be unlocked on reboot

thanks. We probably won't drift from opensource puppet

it’ll take some time to wrap your head around but it should do the trick 🙂

ok, I think I have it. We have a task that takes a hostname and a status message, makes some kind of pretty output and outputs it to a dashboard. for various failure scenarios, we use it to say host:xxx failed the upgrade, or host:xyz - started backups, host abc:backup complete, upgrade starting. that kind of thing. we don't want to wait until all the hosts in a batch are done with their backup before seeing in the dashboard which ones failed (possibly early on in the process). so we end up using parallelize like this:

parallelize(get_targets($correct_version_targets)) |$target| {
  $log_check_results = run_task('projectname::somecomplextask', $target, _catch_errors => true)
  $log_check_results.each |$result| {
    $t = $result.target.name
      if $result.ok {
        run_task('xneelo_pixar::update_status', localhost, {
            server            => $t,
            position          => '99',
            message           => 'done',
            database_password => $database_password
        })
      } else {
        run_task('xneelo_pixar::update_status', localhost, {
            server            => $t,
            position          => '90',
            message           => 'post_checks_failed',
            database_password => $database_password
        })
      }
  }

the idea being that each host will try to do the complex task, then run the dashboard updator as soon as it is done (no matter if successful). Am I right in saying that we could probably get the same result by creating what the earlier example called a subplan, and in that, have 3 tasks. 1. do complex task 2. filter the results of the complex task and run a dashboard updator task on each filter result 3. profit?

Subplan is just a way of organising tasks I’d say.. kind of function

doh! of course