Re: Planned Maintenance - Believe or not ?

From: Tim Gorman <tim_at_evdbt.com>
Date: Fri, 22 Jun 2012 11:46:18 -0600
Message-ID: <4FE4AF6A.7010500_at_evdbt.com>



Management is asking excellent questions in probing the reason for these scheduled restarts. They are probably uncomfortable with IT procedures which are not based on easily-stated fact, and which are likely unnecessary.

You either understand the production environment, or you don't. Scheduled restarts for no stated reason indicate a lack of understanding.

Generally, the term "planned maintenance" refers to the application of patches, upgrades, parts replacement, etc -- in other words, a fix to something known to be broken. Probably the biggest mistake is referring to these scheduled restarts as "maintenance", because nothing is broken and thus nothing is being fixed.

Speculating now, the scheduled restarts probably came about from a specific problem (such as a memory leak), to which regularly-scheduled restarts were implemented as a temporary stop-gap solution. Since the true solution is to fix the root cause, the stop-gap was intended to be used until the permanent fix was implemented, and chances are good the root cause was fixed long ago, but the "temporary" stop-gap solution has persisted beyond the recollection of anyone currently in operations. Perhaps way back then, it wasn't considered burdensome to take things down occasionally, but now availability expectations have likely become higher, and any downtime request is being scrutinized. As they should.

Just my US$0.02...

On 6/22/2012 10:56 AM, Joel.Patterson_at_crowley.com wrote:
> I run single instances, for quite awhile, and with several companies and OS's. For oracle databases, my analogy is that it should be treated like a fighter jet plane. These planes are built to run 24hours/day and operate better that way. They actually require more or completely different maintenance schedule(s) depending on how long they stay idle.
>
> Analogously, oracle doesn't like to span startups, with things like historical reports (Ent mgr), v$ views, and just the extra warm up that happens upon start up to mention some. They do get shutdown once in a while for maintenance reasons, but no schedule.
>
> In the old days, rebooting was done for various reasons. Back in the early 90's, on HP, and Dell, the sysadmins wanted a reboot
> 'the server' on Sunday night believe it or not, (thus the databases went with that). Even running on windows I did not have a scheduled reboot.
>
> Having said that, some people 'keep' a scheduled window for rebooting for ease of administration, (whether for the app or server, or database) -- they always have a window for maintenance and therefore scheduling the maintenance is easier.
>
> Now applications that run in windows boxes sometimes need it, especially with mistakes like memory leaks and the like.
>
>
>
>
> Joel Patterson
> Database Administrator
> 904 727-2546
>
> -----Original Message-----
> From: oracle-l-bounce_at_freelists.org [mailto:oracle-l-bounce_at_freelists.org] On Behalf Of Sreejith S Nair
> Sent: Friday, June 22, 2012 12:34 PM
> To: Oracle - L
> Subject: Planned Maintenance - Believe or not ?
>
> Hi friends,
>
> The question is about the planned maintenance activities you do on a server level or database level. The activity I am referring to is rebooting the node and restarting database instances , crs stack etc. we had this routine from sometime back, but now management asks why should we do it. I have heard from many people that they usually do a server reboot when the uptime goes more than 6 months or so. Now,the management is saying why we have to do it. The plain explanation is that a software is designed to run and to handle a load , then it should run for ever if nothing on it is changed.
>
> I would like to know whether any one has something like a planned maintenance activity where you reboot servers and reboot instances every 6 months or so ?
>
> Regards,
> Sreejith
> -- Sent from my iPhone
> --
> http://www.freelists.org/webpage/oracle-l

--
http://www.freelists.org/webpage/oracle-l
Received on Fri Jun 22 2012 - 12:46:18 CDT

Original text of this message