From: Thomas Northup (thomaslnorthup@yahoo.com)
Date: Thu Apr 21 2005 - 22:21:52 EDT
Thanks for all the replies. They really varied from every week to never (only for patches and kernel changes) The original question was:
***Do you perform downtimes just for the purpose of rebooting the systems?
***Is there a recommended interval HP-UX system should be rebooted?
Below are the answers:
--------------------------------------------------------------------------------------
We have a script that runs and checks the uptime. Then it will reboot the systems that have an uptime of over 60 days on Sunday night. Not actually necessary, but removes stale nfs mounts, zombies, etc. I think the 60 days is pretty arbitrary on our part. Seems to work.
--------------------------------------------------------------------------------------
We try to reboot everything once a month, but sometimes that doesn't happen. I have found that the 11i operating system is quite capable of running basically forever without requiring a reboot. It's the applications on top that need to be restarted occasionally.
--------------------------------------------------------------------------------------
You will probably get a lot of mixed view points so here are my 2 cents. We only reboot serves when they are required, patches, kernel modifications. We do not reboot for the sake of rebooting and any of the shops I have worked in UNIX servers were never rebooted for the sake of rebooting. There are times when we have had to reboot because someone who is not a systems admin, management, requested a reboot thinking it was going to solve a problem. There are the rare occasions when the system is hung and a reboot or a reset is required. The bottom line is I personally do not believe in proactively rebooting UNIX servers.
--------------------------------------------------------------------------------------
Frankly, no. we have had some systems up for over a year without rebooting. The only time a reboot is necessary for us is when certain patch bundles are installed and/or kernel modifications are made.
On the other hand, we have to cycle a handful of applications on the system every so often to flush the memory. This depends on the application, how it is used, and, in many cases, how well it is written. The time involved to do this is minimal compared to a full reboot, but the downtime required for this still needs to be planned.
--------------------------------------------------------------------------------------
As long as your apps are well-behaved, the only reason to reboot is to make kernel patches effective. We do this once every six months. In a few specialized cases where we don't patch, we've had HP-UX boxes up for years without a reboot.
--------------------------------------------------------------------------------------
the time depends on the server model, and also the memory on the servers impacts the time the reboot will take.
i.e. a K is faster than an L, a N is slower, if it has 32 Gb is slower than 4 Gb, etc.
--------------------------------------------------------------------------------------
It depends on the applications that you run on it and how stable they are.
I’ve found in many environments and for most OSs including HP-UX, Windows, Solaris etc, that a weekly reboot keeps things in good working order. We reboot ours on Sunday morning at 4:00am and then another cycle of dependant servers midweek.
--------------------------------------------------------------------------------------
no.
not that i am aware of.
i use both Oracle and Informix along with may other apps in a 24x7x365 environment ...so the goal is no reboots except when forced to add patches, hardware, etc...
--------------------------------------------------------------------------------------
Nope, the only reboots for HP-UX and most UNIX variants are for
patching the OS kernel.
Patching should be done at least quarterly, and more often for systems
with direct Internet connections or other high risk environments.
There is no need to reboot more frequently.
--------------------------------------------------------------------------------------
yes. We reboot ours (most of them) once per month simply for the sake of reboot. For those in a Service Guard cluster it also acts as failover testing.
--------------------------------------------------------------------------------------
Mine stay up all the time unless we have to make a kernel change or hardware change that requires downtime. I've had servers up for 2 years with no problems. It usually works out though that something else requires downtime about once or twice a year.
--------------------------------------------------------------------------------------
In my early admin days I used to think regularly rebooting machines was
a good idea.
Over time my thoughts on the subject have changed. I think you should
only reboot and/or shutdown/power off your systems when necessary. Why
interrupt a service for no *real* reason.
I once looked after a HP-UX box that had an uptime of over 400 days -
pretty impressive. It was almost sad when I had to shut down the
system so an optical tape library attached to it could be serviced.
The adage "if it ain't broke, don't fix it" comes to mind. If a system
is working perfectly ok why reboot?
Now rather than get into a debate of when to reboot, if to reboot, how
long between reboots etc. the better train of thought is this:
How can application services be provided 24 by 7 for every 365 days of
the year while allowing for hardware outages and upgrades?
Well the answer is application failover solutions like HP's
MC/ServiceGuard, CISCO redirector type devices and, ultimately,
building resilience into the application design from day one (great example -
DNS with master and slave servers).
Get the design right in the first place and you can happily run on
hardware that can be frequently rebooted. This also has the advantage of being
tolerant to hardware failure outages.
--------------------------------------------------------------------------------------
>From an OS perspective you should never need to reboot your HP-UX
server. We have some servers that have over 600 days uptime!
About the only reason we reboot a server is to install a kernel patch.
--------------------------------------------------------------------------------------
---------------------------------
Do you Yahoo!?
Yahoo! Small Business - Try our new resources site!
--
---> Please post QUESTIONS and SUMMARIES only!! <---
To subscribe/unsubscribe to this list, contact majordomo@dutchworks.nl
Name: hpux-admin@dutchworks.nl Owner: owner-hpux-admin@dutchworks.nl
Archives: ftp.dutchworks.nl:/pub/digests/hpux-admin (FTP, browse only)
http://www.dutchworks.nl/htbin/hpsysadmin (Web, browse & search)
This archive was generated by hypermail 2.1.7 : Sat Apr 12 2008 - 11:02:47 EDT