Using DrD to patch servers

[likid0]

Hy!,

I use DrD to clone HPIVM, but when it comes to patching/updating I allways use the clasical way of breaking the mirror, patching , and if all goes ok resync mirror.

I am thinking about starting to use Drd to patch and I was wondering how many of you actually use it to patch/update your servers ?

And how you do it?, you break the mirror and use one of the disks for the drd clone?, you add a third(san) disk for the clone ?

there is a nice white paper with info:

http://docs.hp.com/en/5992-1910/5992-1910.pdf

I would like to get some feedback from your experiences with it.

Thnx!.

[Steven E. Protter]

Shalom,

I use Ignite make_tape_recovery to back up vg00 prior to a major patch set and boot off tape to recover if there is a problem.

make_net_recovery is what I use on systems that lack tape drives.

I build my patch sets as follows:
1) Run Security Patch Check and get all patches needed to maintain service contracts.
2) Submit SWA when possible.
3) Get application specific patches
4) Take a current bi-annual update.

Then I combine these all into a single file tape style depot. Then I run one swinstall command and let it boot at completion.

With the newer systems pulling a hot swap drive out of a raid 1 set makes sense. I just would still do the above steps first anyway.

Regards,

SEP

SEP

[likid0]

Yep, I make_net_recovery is allways a must!.

I also put it all together into 1 depot.

What i was curious about if a lot of people are using the new DrD/patch off line system , but it looks like there is not many ;D.

[VK2COT · Accepted solution]

Hello,

I use DRD for all servers in HP Education data centre here in Australia. But I am lucky because I have servers that have
at least three internal disks, so I can
have RAID-1 for the O/S and one disk for
the clone.

Why are not more people using DRD? Here is
my view (based on what I hear from
customers when I teach them):

a) Most of older servers were purchased with
two internal disks only. Extra disk for
clone is not a justifiable purchase as far as IT managers are concerned at the present
time.

b) Most (and I literally mean MOST) customers
have reactive patching policy. In other
words, they apply patches if something
breaks, stops working, or is enforced
by vendors.

Pro-active patching is, sadly, done seldom.
Therefore, having clone boot disk is not such
an important issue because it is used rarely.

c) Lack of familiarity with DRD. Most
customers I teach have never heard of it, or
if they know about it, they have never tried
it.

Hence, we have the "poor advertising campaign" problem :)

d) Humans love proven technologies and
are creatures of habits. Breaking mirrors
has been around for a long time and
Unix admins are comfortable with it.

Having said all this, whenever I
design new servers I try to enforce the
following implementations for cloning boot disks:

AIX (Alternate Root)
HP-UX (Dynamic Root Disk)
Solaris (Live Upgrade)

In fact, my Operation Acceptance Testing scripts look for them:

http://www.circlingcycle.com.au/Unix-sources/

Cheers,

VK2COT

By the way, who am I and what I do:

http://www.circlingcycle.com.au/dusan.html

[likid0]

VK2COT,

Thnx for your answer.

The problem I face is nearly all of my servers have to internal disks in raid1, I have no problem adding a third san disk, but after I install the patches and boot the clone I would be left with a mirror of a san disk and a internal disk, and I am no really into mixing them.

In servers that boot from san, there is no problem with adding a third disk and using drd.

So with servers with 2 internal disks using raid1, I can go with the option of breaking the mirror, patch the clone, boot from the clone and redo the mirror, that's actually ok with me.

[James R. Ferguson]

Hi Daniel:

@ VK2COT : Your comments deserve more than the point system allows. Those are excellent observations.

As for patching and the avoidance of risk at the expense of recovery, Duncan Edmonstone has some excellent guidelines in this thread:

http://forums13.itrc.hp.com/service/forums/questionanswer.do?threadId=1328449

Regards!

...JRF...

[likid0]

Yep, I had actually read that thread, and saved a little note with Duncan comments, great to have with you before being tempted to start patching kamikaze style.

[VK2COT]

Hello,

Firstly, thanks for the kind words about
my previous posting. I appreciate it.

Now to answer Daniels' questions:

a) By default, "drd clone" command
creates a single-disk copy of the O/S.
If that disk is SAN-based (possibly RAID-1,
RAID-5, or similar), there is no need
to have a second disk in "drd clone".

DRD does not use "original" boot disks
(neither primary or alternate).
It uses its own disk(s).

"drd clone -x mirror_disk" option to mirror
the clone when you create it.

b) Like everything in life, good planning
is crucial to success.

I like to spend most of time planning and
verifying. Physical work should be
minimal :)

Best wishes,

VK2COT

[Patrick Wallek]

To VK2COT:

My issue with using DRD to patch is that I don't have any servers running the appropriate OS versions (11i v2 & v3) to allow me to use DRD.

All of my servers are 11.11 or older. Yes, unfortunately still lots of 11.0 and a couple of 10.20.

My single 11.23 server doesn't have sufficient hardware, or the capability for sufficient hardware, to allow me to experiment with DRD.

Alas, maybe one of these days........

[likid0]

I will use DrD, with to disks, to patch some of the preproduction servers next week, and see how it goes.