Significant Maintenance Activity.

[brian_31]

Team:

Can anyone post some significant maintenance activity that was done on HP-UX boxes?

Thanks
Brian.

[Steven E. Protter]

I'm not exactly sure what you mean, but I used to go to Hogwarts.

Every second and forth weekend of the month is a window for me to boot boxes, install patches and do what is needed to keep them running.

Every day we have a batch job that monitors disk space and checks performance.

cron runs security checks of our system with security_patch_check to make sure we are up to date on security patches.

Manually every week we check the security logs and at randome intervals as well.

As far as the physical boxes, we never shut them down or "clean" them unless they are down for other maintenance activity. When HP is doing an upgrade we hand them some compressed air and ask them to blow out the dust.

Is this what you are looking for?

SEP

[John Poff]

Hi,

We just patched our production boxes last Saturday, and we upgraded the firmware on a couple of boxes. Does that count? I swear I didn't break anything, that they know of anyway.

JP

[Robert-Jan Goossens]

Hi Brian,

Only twice a year for production servers, always a complete weekend from friday 19.00 until monday morning 07.30. But we are very strickt about this.

patches + hardware firmware ---> 2 month before production we do the dev boxes.

Robert-Jan.

[A. Clay Stephenson]

I have enough disk drives in arrays or as standalone disks that I routinely replace 1 or 2 drives per week. Because EVERY one of these is eith hot-plug or hot-swap no downtime is incurred.

I also routinely peruse all the patch notices and those that are listed as critical and especially those which contain the phrase "possible data corruption" get my immediate attention. Those might be quickly applied during a brief maintenance window but the normal course is to apply patchsets on a quarterly basis. All of my patches are first applied to a Sandbox, next to a Development/Test environment, and finally Production. The important point is never to exceed your requested maintenance window. Having a multi-tiered test deployment allows you to know exactly how much time is required for the production install. If one always stays within the maintenance window then it is usually not difficult to get them scheduled.

I never routinely shutdown boxes for the sake of rebooting and I assure you that any software vendor who reccommends that practice to me to clear up problems with his software does so only once.

Probably the biggest key to low hardware failure rates is to maintain the datacenter environment. I find that if the temperature, humidity, and cleanliness is maintained along with very stable, clean power that failures other than for mechanical parts like disk/tape drives are all but zero. I am now past four years in our current data center with zero unplanned production downtime.