HP Poll (points rewards): Do you boot different OSes on a same system?

[marie-noelle jeanson]

HP would like to hear from you on this question:
Do you happen to have multiple boot disks on your systems that you would use to boot different OS versions? In other words, do you happen to boot a system from a given disk using a given OS (for instance, HP-UX 11.0) and boot from onother disk (for instance, HP-UX 11i v2)?
Would you need this type of environment in test environment only or any reason why you would need such an environment on production systems too?

THANKS!

Marie.

[Patrick Wallek]

I do not currently do this, but it could definitely come in handy in a test environment.

If a new release comes out it would be nice to install it on a separate disk from previous versions and be able to test. Then, if need be, boot back to previous version of the OS for troubleshooting production issues.

I don't see much of a benefit, in my environment, to booting multiple versions of an OS in production though.

[A. Clay Stephenson]

I certainly do this in my sandbox environment though not in test. I always use a three-tiered deployment: sandbox, test, and production. I currently am able to boot 3 OS versions on my sandboxes. I can't see the need for this in a production environment. I think the question behind the question has to do with virtual boot disks.

[Hein van den Heuvel]

I do, booting either OpenVMS or HPUX on my rx2620 depending on the customer(s) I am working with. Development and test only. No production.

Some day I hope to snarf a Montecito version and run virtual machines to make this whole process more comfortable, always keeping an hpux host in the air, and allocating resources between multiple os'es and/or version of those os'es as needed.
Specifially I want to be able to run more true client-server tests, cross OS, withing my single system.

Btw... my box has 3 drives. I have hpux on 1, opemvms on an other and the 3rd is an hpux data drive. With lvm I carved a up half of those drives (the first half on one, the second hald on the other) and actually use them as data space for OpenVMS. I treat those volumes as 'partitions'. OpenVMS allows me to use a 'virtual disk' driver to present a range of blocks on a spindle as a seperate drive. Not much safety in this setup, but it fills my need. I can test with 3 drives under OpenVMS, with 2 under ux.

Cheers,
Hein.

[Yogeeraj_1]

Hi Marie,

> Do you happen to have multiple boot disks on your systems that you would use to boot different OS versions? In other words, do you happen to boot a system from a given disk using a given OS (for instance, HP-UX 11.0) and boot from onother disk (for instance, HP-UX 11i v2)?
Currently no. we are running hp-ux 11.11. If we upgrade to a higher version, we will most probably do so during our test phase. Unfortunately, since server resources are quite expensive in this region, we cannot keep such a configuration for a long period of time.

> Would you need this type of environment in test environment only or any reason why you would need such an environment on production systems too?
This definitely something very useful. However, as mentioned above, to make efficient use of resources, our test environment are sometimes also used to shared production load.

We would not really need it on the production systems because we would not like to interfere with the availability of systems. Also since we have several servers, such a setup would exists on only one of them.


hope this helps!

kind regards
yogeeraj

[Luk Vandenbussche]

Hi,

We have an rx2620 with the following config

Disk 1 - Partition 1 : W2K3
Disk 1 - Partition 2 : Suse Linux
Disk 2 - Partition 1 : HPUX 11.23

It is used as a test environment.

[Andrew Merritt_2]

I have a test rx2620 which has:

HP-UX 11.23
RedHat Linux (RHEL4 and RHEL3)
SuSe Linux (SLES 9)

boot disks on it. I can't think of a situation other than providing test environments where you'd want to do this.

Of course, on cell based systems you can run multiple partitions, each with a different OS at the same time. On single celled systems, you can do something similar with virtual machines, in some cases (e.g. Linux under Windows, and vice versa.)

Andrew

[Jeff_Traigle]

I've never done this anywhere I've worked, but it could be a benefit for testing in some cases. The only reason I can dream up to do this in a production environment is as a fallback to an OS upgrade rollout since a simple reboot would be a lot quicker than using Ignite.

[Pete Randall]

Marie,

At this point I do not do this, but I have in the past, installing a new OS on a separate boot disk for testing purposes and keeping the prior OS available for fall-back or production. A very handy technique.


Pete

[Heironimus]

We don't do this on a regular basis, but we do keep a spare drive in the machine we use for our internal testing. We have a pretty homogenous environment so the spare drive is normally just the master for our Ignite golden image. If we need to do testing we can swap that drive for a spare and install whatever we need.

We can't upgrade HP-UX right now because of application compatibility, but if we could we'd probably reload that machine first on to new drives and keep the existing OS drives to swap in for testing.

[H.Merijn Brand (procura]

No, I would not.

To be honoust, not even on PC's.

If you want dual boot PC's, my advice would be to use Linux or any other stable *ux OS that supports your VM machine of choice, and run the second(ary) OS in VMware (or the alternative VM machine of choice) on the stable box.

In my work environment, I only have Win2k available on a VMware running on SuSE 10.1 Linux. I never regretted that move. Machine is now up for an avarage of two consequetive months. That used to be 6 days max.

The thing I love about HP-UX is it's stability. Uptime of over a year for 11.00 is quite normal these days.

For small target boxes and test environments, Virtualization is the future if you'd ask me.

Enjoy, Have FUN! H.Merijn

[James R. Ferguson]

Hi Marie:

I don't have multiple boot disks that I would use to boot different OS versions.

I would agree with Andrew Merritt. For servers dedicated to developement or quality assurance testing the mechanics are useful. I can't see the need in pure production environments.

Virtual machines seem closer to modern needs. I make my Windows workstation much more useful by running Linux under EMC VMware coincident with Windows.

Regards!

...JRF...

[John Payne_2]

I don't do this.

I have 8 different Golden Images I use. (Ignite) Every machine I have is a child of one of these images. I have a machine to update these images (or build new ones, as the case may require), which I may install over and over and over, but there is no need for dual boot OS.

A machine is there to serve out it's term with the image it has, when that term is over for whatever reason, we either get rid of the machine, or reinstall it with a different image.

Hope it helps
John

[Geoff Wild]

No - we don't do that.

We may exploit that scenario once we move to VSE on Itanium - but just have different Virtual Servers with different OS levels (11iv2, 11iv3, different patch levels,)

Rgds...Geoff

[Hemmetter]

hello Marie,

we do have one N4000 System with such a multiboot, used for software compatibility tests.
We can boot into HP-UX 11.0, 11i and 11i V2.

Its a pure test environment.

rgds
HGH

[Olivier Masse]

Marie Noëlle,

I've never done that, and if I ever needed to have different OS revisions in the future for tests or development purposes, I agree with others here who would rather use integrity virtual machines for this purpose.

Thanks

[Samir Pujara_1]

Hi Marie,

>>Do you happen to have multiple boot disks on your systems that you would use to boot different OS versions? In other words, do you happen to boot a system from a given disk using a given OS (for instance, HP-UX 11.0) and boot from onother disk (for instance, HP-UX 11i v2)?

Till now I have not done so, but its quit interesting to have such setup in development box.

>>Would you need this type of environment in test environment only or any reason why you would need such an environment on production systems too?

Yes, would like to test it in dev setup and the reason is - I can use same h/w for testing my apps in two different version of o/s. Also it will be very helpful in case of upgradation of o/s.

Suggestion: Any documentation available on this?

Samir

[dirk dierickx]

we don't have or use that kind of scheme here.

[Karsten Löperick]

Hello Marie-Noelle,

I am also not using dual or more OS booting options here. Because I support a more or less pure HP-UX environment. So there is no need for that. On the other hand I do also some checks with Linux. My ( meanwhile ) prefered solution are a VMWare system. That gives me enough flexibility. And as some of the forum members said: the advantage of HP-UX are the stability. So I do not see any necessarity to have multi OS on same system in production environment.

Greetings

Karsten

[Peter Nikitka]

Hi,

I used this in my testing environment. We are using VMware solutions as well, but there are trouble with older OS releases, of course.

In a production environment we used this only at update time as (possible) additional desaster recovery step:
- split mirror
- install new system
- reboot the old state (OS,DB,..) when all other stuff fails

But happily, I never had to perform this step.

mfG Peter

[Tom Henning]

I do not currently do this, but have done this in the past. When I updated some workstations from 10.20 to 11.x, I added a second disk and installed the 11.x on the new disk, keeping the 10.20 disk as a quick rollback fix. I could very easily see doing this in OS version upgrades on the test box, but would have no need for this in production.

[Jan van den Ende]

marie-noelle,

>>>
Do you happen to have multiple boot disks on your systems that you would use to boot different OS versions?
<<<

Well YES & NO

We USED to, when the cluster existed of Vax and Alpha machines (and probably will again when we start using I64) because each architecture MUST have its own system disk.

Also, we USED to before we started using multi-site SAN, because it is unsupported to NOT have a site-local systemdisk (Although it IS possible to shadow a system disk over sites, there are some "interesting" issues when one total site is down and then boots from the a local drive, that is not currently member of the active cluster system disk shadow set).

We regularly (about once or twice a year) do USE it during rolling upgrades, to go forward on OS version or patch level.
We also did use it during some of the node replacements, but mainly because that provided a nice opportunity to catch up on patch level.

>>>
Would you need this type of environment in test environment only or any reason why you would need such an environment on production systems too?
<<<
Au contraire: we can well accept some downtime for our test system, but production has a strict 365.25 * 24 uptime need, so any updates MUST be done rolling.

hth,

Proost.

Have one on me.

jpe

[Andrew Young_2]

Hi Marie

We use alternate (cloned) bootdisks when testing OS upgrades or patches we think are "risky". However we are now looking to do our testing and development using Itanium Virtual Server Engines (VSE's). However hardware patches will still be tested on cloned disks.

Note with VSE's - don't try any performance computing on them, IO was really slow on the Madison's, but will be playing with the Montecito's next month which I hope will be better.

We currently have no need for running different OS's in production.

[Bob Ingersoll]

I have done this both on PA and Itanium platforms for testing and OS upgrade purposes on production servers.

[siva0123]

Hi,

Though i'vent used or tested this feature in HPUX , i feel this should be a nice one to have in production boxes so that we have a live machine which doesnt affect production whenever there is an upgrade or other ones and just a reboot is what required to switch over the production box to newer version.

I presume installing a seperate os version in an alternate disk doesent require a down time as there is an option with one of the other UNIX flavor for this.


Thanks,
Siva