November 11, 2004 at 6:18 am
Hello everyone,
Is it a good practice (if we have the time frame to do it) to reboot the database server on a weekly basis?
This is something we are doing for years here. When our db server were on NT 4.0, the stability on the os side was not always there.
But now with Win 2K and Win 2003 is it a useless practice?
Maybe it could be done more on a monthly basis or a quaterly basis.
What are your comments, suggestion experience about that?
Best Regards,
Carl
November 11, 2004 at 6:37 am
why would you want to reboot the server anyway ? I sort of understand that you might want to check the failover of a cluster but to generally reboot a sql server would be a bad thing as the cache(s) would be cleared and performance degraded for a day or so.
[font="Comic Sans MS"]The GrumpyOldDBA[/font]
www.grumpyolddba.co.uk
http://sqlblogcasts.com/blogs/grumpyolddba/
November 11, 2004 at 6:48 am
In the late 90, when we were using NT4.0 as the O.S. on the database server, we were doing that for the stability (I guess because I was not here). I knew that we encounter serveral server crash and when they decided to reboot the server on a weekly basis, they gain more stability on the OS side (less crash).
Maybe, as I wrote it, today with WIn 2K and Win 2003 this practice is completely useless?
It is not to check the failover of a cluster.
November 11, 2004 at 6:54 am
Interesting, I ran NT4 ent servers for periods of over 6 months without problems.
I generally don't reboot servers unless forced.
[font="Comic Sans MS"]The GrumpyOldDBA[/font]
www.grumpyolddba.co.uk
http://sqlblogcasts.com/blogs/grumpyolddba/
November 11, 2004 at 6:56 am
Thank's for your intput Collin,
Carl
November 11, 2004 at 11:27 am
Since there have been no patches for SQL Server in quite a while some servers have been up for 6 months or more. About the only reason they get rebooted is for a Windows patch.... We are on all Win2000.
November 11, 2004 at 12:28 pm
I agree with Colin....
Not only the cashes are empied....
but think about the downtime (if this is production server)
think about dependencies.
think about extra failures when not comming up properly (its still a MS product !!)
think about extra time you have to work (in your standby shift (mostely at night)).
etc etc.
We have serveral servers running for a long periode which only needs theire OS reboot after patching.
Be sure you run several maintenace jobs ( Reorg. Reindex etc ) on a frequentely base this prevents slowing down performance and keep up the stability.
GKramer
The Netherlands
November 12, 2004 at 4:01 am
Several Companies would schedule weekly reboots in the NT OS to overcome the memory leak issues. Have enough physical RAM to outlast the "leak". Then clear all of the reserver,buffers, etc... through the use of the weekly reboot.
2000 and 2003 OS's eliminated almost all of these issues. (I have had production machines up for over a year. I had to down it to patch).
If you are concerned, Check system statistics and make the decision. Enough patches are coming out to force once a quarter downtime any way.
Cheers,
Zort2001
November 12, 2004 at 10:30 am
I agree with the no reboot policy. My production server runs for months with no attention or reboots. Server 2003 and SQL 2K are rock solid.
So long, and thanks for all the fish,
Russell Shilling, MCDBA, MCSA 2K3, MCSE 2K3
November 12, 2004 at 10:34 am
Thanks to all of you.
Carl
Viewing 10 posts - 1 through 9 (of 9 total)
You must be logged in to reply to this topic. Login to reply