On Thu, Jan 8, 2009 at 6:43 PM, Kory Wheatley <[email protected]> wrote:
> I would like to know other system admins patch strategy for
> installing Solaris 10 Patches.  Typically, we install the Sun Alert
> Security Patches on our test systems then try to roll them out
> within 30 days of the patch coming out.  We only do recommended
> patches when needed by software on the system or if required as a
> dependency patch

Another mechanism to consider is requesting that Sun run an SRAS audit
against a representative system.  While security is an important
aspect of system availability, the recommendations that come from an
SRAS audit (in particular EIS analysis) can be useful.  This has the
most value on systems that fall out of the range covered by the
Recommended cluster.  For example, for years neither the Sun Alert
bundle nor the Recommended bundle had patches for ge or ce network
drivers.  However, when network performance problems were detected
patching the network drivers was almost always the first step toward
resolution.  Throw in some high end (e.g. E25K, M8000) servers or SAN
drivers and you will get even further from the ground covered by the
Sun Alert and/or Recommended bundle.

> Also do you break your mirror to install the patches on one disk and
> test than sync back up (either using Solaris Volume Manager or a
> hardware mirror)?  Do you only do this for Kernel patches and
> Solaris 10 Update patches?    It would seem time consuming breaking
> the mirror to install all security patches, but I don't know what
> others are doing.  We recently don't use this method.

Live upgrade.  The EIS boot disk layout[1] is the mechanism that I
have used for years - even before it was Sun's recommendation.  It
works well.  If you are doing a split mirror, be sure that you make
the split mirror bootable before you need it.  That is, split, mount
at /mnt, metaroot -R /mnt, umount /mnt, THEN patch.

1. http://wikis.sun.com/display/BigAdmin/BootDiskLayout

With ZFS root, Live Upgrade becomes much cooler.  Creating a new fully
functional boot environment takes seconds (if you include the time
required to type "lucreate" with a couple options) and takes up a
trivial amount of space.  Of course, if you put a gig of patches in
it, it will grow.  However, it makes it feasible to have many boot
environments in the space previously required to store only two.

> I would just like to get a feel if other System Admins are
> aggressive in installing the security patches as soon as they come
> out?  Versus why break it if it runs.

My preferred approach is to harden the system such that it is only the
very occasional network-based attack on ssh or other must-have
services that will have a high likelihood of being exploited.
Establish good identity management practices so that only those that
should have access to the system have it and they are all done through
named (not shared) accounts.  Lock out people that shouldn't have
access as soon as their need for access ends.  With that, the need to
apply the latest patches (for e.g. Java applet vulnerabilities, again)
tend to not look very important.  Secure by Default and JASS can be
your friends here.

Since you are coming from a .edu, the case where you have hundreds or
thousands of students accessing as system needs to be covered as well.
In that case, patching the locally exploitable bugs becomes much more
important but will never be a complete solution.  Without policies in
place that give you a way to get rid of authorized users that
misbehave, you have little chance of having a long-term secure system.
Get to know the relevant computer use and misconduct policies and work
with those that can help you enforce them.

One interesting approach suggested by Bob Netherton is to have three
boot environments at all times.

- The previous "good" boot environment
- The current running boot environment
- The next boot environment

Whenever a patch has passed your QA process, apply it to the "next
boot environment".  That is not to say that you activate it at that
time.  This prepares you for always being ready to activate it when
the outage window comes, or when threat level associated with the
patches you have already applied (but not activated) suddenly
increases.  At such a time as you activate your next boot environment
and you are happy with it, destroy the oldest boot environment and
create your new next boot environment.

-- 
Mike Gerdts
http://mgerdts.blogspot.com/
_______________________________________________
sysadmin-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/sysadmin-discuss

Reply via email to