I replied over on the WISPA list, but here it is again:

Scheduling a 'yum -y update' is really not a good idea.

Updates can be tricky, but in larger shops, I often use Pulp to manage yum
repositories for fleets of RHEL/CentOS boxes. I sync RHEL/CentOS
repositories to my Pulp repos nightly, and then take snapshots of these
repos for "non-production" and "production" hosts.  The servers themselves
point to the 'snapshot' repositories (non-production hosts point to the
non-production snapshot, production hosts point to the production
snapshot).  This allows me to promote bleeding edge/new packages to
non-production hosts and test them before promoting to production hosts.

This also gives me the ability to cherry-pick very important updates (eg,
critical security updates) and quickly promote them to all hosts.  From
there, I typically use Puppet/mcollective to trigger "yum updates" on the
remote hosts.  You can also use the pulp-consumer client which would be
installed on each host and would allow you to "push" updates out to the
hosts from your central Pulp server.

Again - this may be overkill for smaller shops.  There are several
'home-grown' things you can do as well (using SSH to execute simultaneous
yum updates across environments, etc) that may fit your environment better.

Josh

On Wed, Jan 7, 2015 at 9:38 AM, Mike Hammett <[email protected]> wrote:

> In Windows, there's WSUS, SCCM, etc. for managing the update process on a
> bunch of machines. How are you guys managing that in Linux? I have mostly
> (if not completely) CentOS and Debian, ranging from version 5 (I hope not
> any older) to version 7.
>
> I'd like to think that I'd want something more elegant than a cron that
> does yum update -y daily or weekly, but really, I never check the packages
> I do upgrade, I just upgrade them all.
>
>
>
> -----
> Mike Hammett
> Intelligent Computing Solutions
> http://www.ics-il.com
>

Reply via email to