Puppet Community

Usually, `bolt` spends most of its time parked in `read()` waiting for something over the network to return.

It can be memory intensive if you run a command or task that generates a lot of output across a large group of targets.

Usually not CPU intensive, but I could definitely write a custom function that hammered CPU if I wanted to :wink:

I'm using it to deploy a command to a couple thousand nodes whenever necessary.  I'll bump up the RAM then.  Appreciate the info!

I had to add ulimit -n in my bash profile

speaking of.  Is there a pattern or reason behind how bolt goes about connecting and sending commands?  Like ansible has a few different methods for working it's way through an inventory.  Does bolt have logic like that or is it just shotgun blast it's way through?

you have options to do batches. we don use :ansible:-word here :slightly_smiling_face:

also you can write a plan that will do targets sequentially instead of in parallel, if that's what you need

sequentially would probably be the best way for me to approach it

I think bolt by default only does 50 or 100 targets in parallel?

I didn't know that. Thank you for the information <@U02C3FKFTU7>