Something about this box causes Salt to be extremely slow

[chuyskywalker]

I was attempting to use Salt to provision a machine based on this box, but as you can see in this gist a basic salt state with nothing in it takes over 60 seconds to run.

If I spin up the same salt provision, but with this box:

  config.vm.box = "puppet-labs-centos64"
  config.vm.box_url = "http://puppet-vagrant-boxes.puppetlabs.com/centos-64-x64-vbox4210-nocm.box"

the salt highstate takes about 2 seconds

[casr]

Thanks for the report. Unfortunately, I'm not familiar with Salt.

Is it perhaps a missing dependency that takes time to download?

Another route that might be worth testing would the 6.4 box and seeing if you have the same problem.

Does Salt have high memory requirements? Perhaps GH-8 might be related in that case.

[chuyskywalker]

As it takes 60 seconds every run (not just the first) it's probably not some dependency like a first-run would have. Plus, as noted on another centos box it runs super fast, every on the first run.

It could be memory, but I give the VM 2GB of ram at boot, so probably not. I mean, it shouldn't be swapping by that point anyway.

...and I just tried the older box image your linked to, and we have a winner: it's not having the same problem. With https://github.com/2creatives/vagrant-centos/releases/download/v6.5.2/centos65-x86_64-20131219.box It takes the weird 60 seconds to run a basic "echo hello" provision. But that same salt provision in https://github.com/2creatives/vagrant-centos/releases/download/v6.4.1/centos64-x86_64-20131218.box works in 1.3 seconds.

[chuyskywalker]

It's worth noting that, from the 6.4 image, I could do a yum update to get to 6.5 and the salt call still takes about 2 seconds. So something about the newer box is probably behind this.

[casr]

Seems like your problem might be related to CentOS 6.5? Just to be sure you could try the older releases of 6.5 on the release page or perhaps vagrantbox.es has some 6.5 boxes too (might have to look in the pull requests as the repo isn't updated often).

[casr]

It's worth noting that, from the 6.4 image, I could do a yum update to get to 6.5 and the salt call still takes about 2 seconds. So something about the newer box is probably behind this.

Missed this. That is interesting...

Does it still go fast if you yum update before installing Salt?

[chuyskywalker]

So I've tried a few of the older 6.5 boxes:

build	real	user	sys
v6.5.0/centos65-x86_64-20131202.box	0m1.318s	0m1.196s	0m0.074s
v6.5.1/centos65-x86_64-20131205.box	0m1.605s	0m1.234s	0m0.099s
v6.5.2/centos65-x86_64-20131219.box	0m1.395s	0m1.244s	0m0.083s

I...can't replicate this now. Damn hiesenbugs. Thanks for giving me some pointers anyway. If I can produce a reliable replication, I'll send over some better instructions.

[casr]

D'oh! Good luck with it!

[calvinhp]

I was experiencing the same issues with your 6.5 box. I switched to the puppetlabs 6.5 box and the issues went away. During the salt runs it would sit in wait for a long time before actually executing any of the python code from what I saw, but I'm not certain what is different between the box builds.

[casr]

What's the date on the centos65 box are you using? Could you try some of the other dated releases off of the releases page and see if that helps?

[casr]

@calvinhp Just looking through sshd_config, it probably relates to 5b79451. The 16 Jan release would have had it under the "Subsystem sftp" line thereby making it ineffectual, I believe.

I will have to package up a new release soon, I guess!

[calvinhp]

That would make sense, I look forward to the latest draft...

[casr]

Might also be related to IPv6. hashicorp/vagrant/issues/1172