So I've got a large inventory set I'm working with, on the order of 20K entries via a plugin.
I'm doing some testing with just one host in use, so a --limit down to a single host.
I've created three identical plays in a book, which just do a command: sleep 1.
With 1.2.3 this execution would take 30~s
Same with 1.3.4
with 1.4.3 this goes up to 1m40s
Has something changed with what ansible does between plays? I hadn't noticed this at first as I was working with one of my much smaller inventories.
Can anybody verify what I'm seeing? If so I'll happily file an issue.
-jlk
Lots have changed all over, but we don’t have access at all to your environment to know where to start.
You have an inventory script in play, I think, but we don’t know the particular parameters of it.
It would probably best if you can identify what is thinking about things on your end through some profiling.
I'm trying to bisect between 1.3.4 and 1.4.0 to find where things start going downhill.
As for inventory, yes it's a inventory script, with maybe 12 high level groups which each hold 10~20 sub-groups each, with hosts being in those subgroups. Variables at the sub-group level, and a tiny amount of hostvars defined via _meta.
Doesn't matter where the action happens, could be a local task, there is still a long pause after a the task completes before the next play header is printed (18~seconds), and a shorter pause between the play header and the actual task header (5~seconds).
I haven't been able to easily capture what's going on with an strace yet, I'll keep working on it.
-jlk