[jldugger]

We actually have our Chef rdiff backup cookbook randomly distribute jobs across a buckets of time using a hash function of the hostname.

[sillysaurus3]

I have to know: Why a hash function of the hostname?

[pliu]

The chef-client cookbook does a similar thing in its cron recipe:

  # Generate a uniformly distributed unique number to sleep.
  if node['chef_client']['splay'].to_i > 0
    checksum   = Digest::MD5.hexdigest(node['fqdn'] || 'unknown-hostname')
    sleep_time = checksum.to_s.hex % node['chef_client']['splay'].to_i
  else
    sleep_time = nil
  end

https://github.com/opscode-cookbooks/chef-client/blob/master...

This is random enough so you won't kill the server, and deterministic so the resource isn't always changing every Chef run.

[toomuchtodo]

No hash collisions, hostnames (in almost all practical environments) are never identical.

[jldugger]

It's more or less random, but stable.