novel

at the end he's just like 'dont use windows lol'

hi 👋

o/

Depending on the actual impact of pluginsync'ing providers, there are different solutions. None of the solutions are to disable pluginsync since that's not really a solution... pluginsync is a good thing, though it has some sharp edges. If the problem is longer than expected initial provisioning times, pre-loading the plugins as part of the puppet agent installation is a good strategy. That's what Puppet Enterprise does. The PE agent installer script just fetches an archive of the plugindir contents and unpacks it before doing the first Puppet run. Or just build an image that has some of the plugins pre-downloaded. If the problem is excess load on the Puppetserver affecting regular Puppet runs and bringing down the whole system, I've had great success with putting an nginx reverse caching proxy in front of the Puppetservers to cache all the file serving requests. That works because Puppetserver is terrible at serving files, but nginx is amazing at it. If the problem is neither of those things, and it's just annoying to look at, then don't bother trying to fix it because It's not worth the effort.

interesting info. will do some research on that. was more of an investigative/optimization thought. on 15s puppetagent time, nearly 1/3 is spent on pluginsync. the nodes manages ~200-500 resources. removing some non used stuff definitely helped, but maybe i just maxed out the optimization.

I don't think that's bad

i am using single beefy docker container, on 10g networks, with nfs.

just manage more stuff with windows on the node, so the agent runs longer

lol

makes the relative time spend for pluginsync shorter!

so on larger environments, my customers have 1500-6000 managed resources per node and a puppet run takes like 75-900 seconds

spending 5s to check if all plugins are up2date, isn't really a problem

Ah yeah, 5s for plugins and prefetching is minor. Not a problem that I would bother trying to solve. Similarly to Bastelfreak, pluginsync was a problem worth solving when I had to account for 10's of thousands of nodes with 60s+ runs that started ballooning to multiple minutes per run.

yeah. I would really love to see the nginx, that's already part of PE, caching all file resources and pluginsyncs 😞

there's even a ticket for it, in the now-ignored jira

Oh, did PE implement file caching?

lol

no

all cool tickets get ignored

ah, misread.... yeah, there's a ticket, that's it 🙂

I'm trying to modify some settings of puppet-lint. I wanted to permanently exclude one of the checks that is being run, is there anyways to do this via a config setting? (I'm running it from a Windows machine, if that makes any difference).

I'm on windows, so I'm not sure if the puppet-lint.rc file is stored in a different directory. I don't see it anywhere in my home directory.

well, try to create it and see if it works

won't exist on it's own. You'd need to create it. IIRC the puppet-list binary looks for a .puppet-lint.rc locally and then the dirs you've pointed it at.

Awesome. That worked. Thanks guys 🙂

I need a help, I have a filebeat module. Now we are removing filebeat from all the servers. Using the new module version in puppet file I am able to remove,stop filebeat and related files, but still show the filebeat process

Anyone running a puppet server in GCP, and have documentation on how to set it up properly?

I guess the answer is “no”, but still.. Let’s say I have a custom Pattern[/^(foo|bar)$/] type. Now I want to make another type, which can handle multiple comma-separated entries of

foo

or

bar

. Is there any chance that it’s possible to include/merge multiple Pattern types somehow maybe?