Hi,

Thanks in advance for reading this. Here's what I've been doing thus far:

* SVN server with several repos whose contents should be available to all systems
* NFS server with read-only export (that directory contains the various SVN working copies that I want all systems to have access to)
* 30+ other servers with a NFS mount to that NFS servers's export
* Symlinks from that NFS mount point to various locations

The symlinks are to common scripts and config files.

Examples:

* Interactive script to remove old kernels
* Cron script to check for and report patches
* Interactive script to VMware Tools (admittedly it is short)
* logrotate conf files
* Cron script to call Tivoli client to backup specified directories (the list of directories is unique to each system while the script itself is written to run on all systems)

I've been debating going with a different approach:

* SVN server (same as above)
* All servers with individual working copies
* All servers with a cron job that updates the working copies

My main goals are reliability and ease of maintenance. I've had several systems that have "lost" NFS mounts and failed to restore the connections. This led to various cron scripts that did not execute because they couldn't be accessed. I've started to look into a cron script (or Nagios) to monitor NFS mounts and even AutoFS (of which I had little luck), but I came to the realization that I may be making this more complex than I need to.

Between those two approaches (SVN working copies shared via NFS or local SVN working copies), which strikes you as the most reliable? I'm leaning towards local working copies at this point.

Thanks for your time.

The individual working copies scenarios sounds simpler (based on the assumptions in my head ) - there are fewer components, just the svn working copy and the script to update it.

Just a couple of points:

For your NFS scenario, can you just test for the existence of a known file before your script executes?

A drawback of the local working copy would be the lag between changes being made and migrated out to the other working copies. Is that a risk for you?

Thanks for the reply gilead.

Basically a script calling another script, with the parent script verifying that NFS is available? I thought of having a separate cron job that would do something like that and force a remount of the NFS volume if not found. Sounds like a great use of Nagios (which I'm not familiar enough with yet).

Not in our setup. Right now I'm the main person making changes to the repos, and the changes I make are often minor ones. That's one reason I'm leaning towards SVN working copies on each of the boxes + a cron job to sync them. Once I do get more familiar with Nagios it could probably run a script to bring the repos into sync if I wanted something more advanced.

NFS is nice when it works though, as everything is instantly up to date & managed from one location.