Your box uptime?

Basis (Basis Technology Modules: Basis Component/System Administration, GUIs)

Moderators: Snowy, thx4allthefish

Post Reply
blue_teeth

Your box uptime?

Post by blue_teeth » Wed Aug 27, 2003 10:04 am

Hi,

Perhaps this is the most impertinent question asked on this forum.

What is your current Production SAP/Unix box uptime?

Ours is 338 days!

Last we bounced was to test HA failover.

Regards,
BT

bubu
Posts: 529
Joined: Mon Oct 21, 2002 7:42 am

Post by bubu » Wed Aug 27, 2003 11:04 am

Just wondering if there are any Windows system there running any SAP system (e.g. R/3, CRM, BW, etc) that have almost the same uptime ...

Regards,

dingbat
Posts: 141
Joined: Sun Oct 27, 2002 7:36 am
Location: UK

Post by dingbat » Wed Aug 27, 2003 11:28 am

bubu wrote:Just wondering if there are any Windows system there running any SAP system (e.g. R/3, CRM, BW, etc) that have almost the same uptime ...
you mean measured in hours instead of days...?? :lol: :lol: :lol: :lol:

(sorry, I couldn't resist... I'm sure there are...)

dingbat.

oldsapguru
Posts: 5489
Joined: Wed Oct 23, 2002 6:08 pm
Location: Nashville via Manhattan
Contact:

Post by oldsapguru » Wed Aug 27, 2003 11:31 am

My UNIX box gets bounced when something bad happens to our Dallas center. Otherwise, it never gets cycled. Our NT server, however, gets cycled once a week or so.

Guest

Post by Guest » Wed Aug 27, 2003 1:32 pm

oldsapguru wrote:Our NT server, however, gets cycled once a week or so.
Seems to be necessary if you with Windows systems.
Even SAP is rebooting the OSS system once per week to "improve performance"

oldsapguru
Posts: 5489
Joined: Wed Oct 23, 2002 6:08 pm
Location: Nashville via Manhattan
Contact:

Post by oldsapguru » Wed Aug 27, 2003 3:33 pm

Guest, funny how that works, huh?

MichaelPhillips
Posts: 554
Joined: Mon Oct 21, 2002 10:05 am
Location: MEMPHIS - TN

Post by MichaelPhillips » Wed Aug 27, 2003 4:49 pm

mine is 372 days.......... last time was also for HA testing....

myBASIS
Posts: 96
Joined: Thu Jul 17, 2003 10:25 pm

Post by myBASIS » Wed Aug 27, 2003 8:48 pm

My servers as per customer policy will be rebooted once everymonth. So uptime never go beyond 31 days. But I really do not know how good is it rebooting SAP HP-UX servers once everymonth. Any comments....
myBASIS

oldsapguru
Posts: 5489
Joined: Wed Oct 23, 2002 6:08 pm
Location: Nashville via Manhattan
Contact:

Post by oldsapguru » Wed Aug 27, 2003 9:08 pm

If you aren't fighting HA and you have the time, it doesn't hurt.

Dman
Posts: 264
Joined: Wed Apr 02, 2003 9:56 am

Post by Dman » Thu Aug 28, 2003 9:33 am

Looks like most of us are in the same boat.
I run on Unix and i also only have to reboot for HA failover tests,
and performance is great.

if thats not tempting fate i dont know what is !!!!! :twisted:

Daz.
You only live once, live it right,
and once is enough

blue_teeth

Post by blue_teeth » Thu Aug 28, 2003 9:54 am

My servers as per customer policy will be rebooted once everymonth. So uptime never go beyond 31 days. But I really do not know how good is it rebooting SAP HP-UX servers once everymonth. Any comments....
There are people who argue that systems be bounced at regular intervals (say once in one/two months) so that memory fragmentation is taken care of. Perhaps they come from Microsoft camp?

Now, on Unix platforms, my experience is this:

1. Rarely, if ever, the memory gets fragmented and requires bouncing.

2. When rebooted, it takes awhile (2-3 days) for system to become "warm" taking performance hits.

Now, I'm not conversant with HP-UX and it's underlying hardware. However, my experience with SunOS, AIX and Tru64 shows rebooting is only necessary when new OS kernel patches are applied or somesuch stuff. Memory fragmentation and system performance degradation due to long running? No Sire!

Regards,
BT

Guest

Post by Guest » Thu Aug 28, 2003 9:55 am

I am running on Windows (used to run on AIX and miss it deeply). I am now bouncing the boxes once per month. We tend to have less mysterious issues. The first thing taguht at a MCSE class is to re-boot because it solves most Windows problems. The second thing they teach is to re-boot again just to be sure. The third thing they teach is to call Microsoft to get random hot fixes they do not pertain to your problem one bit but they must be installed to stay in support (or lack of support).

oldsapguru
Posts: 5489
Joined: Wed Oct 23, 2002 6:08 pm
Location: Nashville via Manhattan
Contact:

Post by oldsapguru » Thu Aug 28, 2003 11:00 am

I have had to reboot an AIX box due to stuck ipc chunks. That puppy was blowing chunks everywehere! The ipcs wasn't showing anything but I was still getting shm errors. Reboot cleaned it all up.

Blue_Teeth is correct, however. Once you cycle a AIX box, it takes awhile to figure out what should stay in the buffers, etc. The suckers also "go to sleep". If they aren't being used for 12 hours or so, they go to sleep and it can take a good bit of time to wake them up and convince them to stay awake. Its sorta like it starts the rewarm process all over again. I'm sure this has to do with active task polling.

bubu
Posts: 529
Joined: Mon Oct 21, 2002 7:42 am

Post by bubu » Thu Aug 28, 2003 12:14 pm

Now, I'm not conversant with HP-UX and it's underlying hardware. However, my experience with SunOS, AIX and Tru64 shows rebooting is only necessary when new OS kernel patches are applied or somesuch stuff. Memory fragmentation and system performance degradation due to long running? No Sire!
Same is true for HP-UX. My previous employer has us as our policy to have a sheduled downtime for the server only during maintenance period with the vendor which normally occurs ever six months.

Regards,

myBASIS
Posts: 96
Joined: Thu Jul 17, 2003 10:25 pm

Post by myBASIS » Thu Aug 28, 2003 9:39 pm

Thanks to all.

I totally agree that rebooting Microsoft WINNT and W2K servers is needed as a policy once in a month(or peroidically). I also agree that Unix(all flavors) do not require reboot as a policy perodically. I want more info on the good and bad part of having it as policy for unix servers. This will help me to initiate change policy (for good of course).

The immediate good reasons I could see is: Performance issue on the first working day after reboot due to direct db access resulting more response time than normal. Although this doesn't really noticed by user due to little more milli seconds, a point to acknowledge when it comes to response time.

Regards
myBASIS

Post Reply