"A Freudian slip is when you say one thing, but mean your mother."—Anon
We all make mistakes; that's why pencils have erasers. Whenever Puppet changes a file on the client, it keeps a backup copy of the previous version. We can see this process in action if we make a change, however small, to an existing file:
# puppet agent --testinfo: Caching catalog for cookbookinfo: Applying configuration version '1293459139'--- /etc/sudoers 2010-12-27 07:12:20.421896753 -0700+++ /tmp/puppet-file20101227-1927-13hjvy6-0 2010-12-27 07:13:21.645702932 -0700@@ -12,7 +12,7 @@ # User alias specification-User_Alias SYSOPS = john+User_Alias SYSOPS = john,bobinfo: FileBucket adding /etc/sudoers as {md5}c07d0aa2d43d58ea7b5c5307f532a0b1info: /Stage[main]/Admin::Sudoers/File[/etc/sudoers]: Filebucketed /etc/sudoers to puppet with sum c07d0aa2d43d58ea7b5c5307f532a0b1notice: /Stage[main]/Admin::Sudoers/File[/etc/sudoers]/content: content changed '{md5}c07d0aa2d43d58ea7b5c5307f532a0b1' to '{md5}0d218c16bd31206e312c885884fa947d'notice: Finished catalog run in 0.45 seconds
The part we're interested in is this line:
info: /Stage[main]/Admin::Sudoers/File[/etc/sudoers]: Filebucketed /etc/sudoers to puppet with sum c07d0aa2d43d58ea7b5c5307f532a0b1
Puppet creates an MD5 hash of the file's contents and uses this to create a filebucket path, based on the first few characters of the hash. The filebucket is where Puppet keeps backup copies of any files that it replaces, and it's located by default in /var/lib/puppet/clientbucket.
# ls /var/lib/puppet/clientbucket/c/0/7/d/0/a/a/2/c07d0aa2d43d58ea7b5c5307f532a0b1contents paths
As you just saw, the # ls command listed the filenames. You will see two files in the bucket location: contents and paths. The contents file contains, as you might expect, the original contents of the file. The paths file contains its original path.
It's easy to find the file if you know its content hash (as we did in this case). If you don't, it's helpful to create a table of contents of the whole filebucket by building an index file.
How to do it...
Create the index file using the following command:
The script will create a complete list of files in the filebucket, showing the original name of the file, the bucket path, and the modification date (in case you need to retrieve one of several previous versions of the file). Once you know the bucket path, then you can copy the file back into place.
There's more...
You can have Puppet create backup copies of the file in its original location, rather than in the filebucket. To do this, use the backup parameter in your manifest:
Now, if Puppet replaces the file, it will create a backup version in the same location with the extension .bak. To make this the default policy for all files, use:
File {
backup => ".bak",
}
To disable backups altogether, use the following code: