Re: database activity when mounted
- From: Jozsef Horvath <horvathjozsef@xxxxxxxxx>
- To: Josh.Collier@xxxxxxxxxxxx, ORACLE-L <oracle-l@xxxxxxxxxxxxx>
- Date: Fri, 5 Nov 2010 11:17:04 +0000
Josh,
Since I do not know Netapp nor MSO much, I can't really have idea why it
works like this. Having said that I find it really strange that a tool
performing a database restore mounts the instance (makes sense: gain access
to the control file) then tries to unmount the filesystem underneath.
My first thought: misconfiguration or Netapp bug.
Does the documentation confirms this is the expected behaviour? (mount the
instance, unmount the FS)
- Jozsef
On 4 November 2010 17:01, Josh Collier <Josh.Collier@xxxxxxxxxxxx> wrote:
> The tool that I am using to do the snapshot restore (netapp snap manager
> for oracle) places the database in a mounted state before it tries to do the
> snaprestore. I have no idea why, nor am I able to control this
> functionality.
>
>
>
> I agree, if it was me and I was in control. I would shutdown everything and
> then lay the snapshot in.
>
>
>
> I have tried removing the init file so that its impossible for the SMO to
> start the database but if fails.
>
>
>
> It really looks to me like there isn’t any activity on the volumnes and the
> error is a VxFS or netapp issue.
>
>
>
> Thanks for your time,
>
>
>
> Josh C.
>
>
>
> *From:* Jozsef Horvath [mailto:horvathjozsef@xxxxxxxxx]
> *Sent:* Thursday, November 04, 2010 9:34 AM
> *To:* Josh Collier
> *Cc:* Hemant Chitale; ORACLE-L
>
> *Subject:* Re: database activity when mounted
>
>
>
> Hi Josh,
>
> I am probably not the only one who is wondering why would you want to "pull
> out" any device under the running instance, even it is in mounted state?
> I appreciate that I am not aware of you business needs but maybe if you
> could share more information about your environment, the requirements then
> you may have receive more, constructive response.
>
> The way I see it: if I do need to do any maintenance work on the server
> where there is/are oracle instance(s) I make sure it can be done while the
> instance is up&running. I case it is not possible and the SLA allows
> downtime then I shut the database down beforehand do the maintenance work.
> Though, it makes me thinking that if your instance is in "only" mounted
> state what could prevent you to shut it down properly then unmount the FS
> underneath? I suppose it is not a standby instance otherwise you would
> anticipate activities on the volumes.
> Again, important information may be missing here.
>
> Regards,
> Jozsef
>
> On 4 November 2010 16:19, Josh Collier <Josh.Collier@xxxxxxxxxxxx> wrote:
>
> No responses.
>
>
>
>
>
> I have not tried an os trace. That is a good idea.
>
>
>
> Let me ask you this.
>
>
>
> Do you know of any other way besides fuser to detect activity on a volume?
>
>
>
> *From:* Hemant Chitale [mailto:hemantkchitale@xxxxxxxxx]
> *Sent:* Thursday, November 04, 2010 1:38 AM
> *To:* Josh Collier
> *Cc:* ORACLE-L
> *Subject:* Re: database activity when mounted
>
>
>
> Did you get any responses? Have you done an OS trace of DBWR?
>
> Hemant K Chitale
> http://hemantoracledba.blogspot.com
> sent from my smartphone
>
> On Nov 2, 2010 6:48 AM, "Josh Collier" <Josh.Collier@xxxxxxxxxxxx> wrote:
>
> I have a question for you.
>
>
>
> What would cause activity on a filesystem that contains only datafiles when
> the database is mounted?
>
>
>
> My database is segregated so that control and redo files are isolated from
> datafiles and tempfiles.
>
>
>
>
>
>
>
> Normally what I expect is this
>
>
>
> 1. Database is in mounted state
>
> 2. Ckpt process is active on the mount that contains the control
> files
>
> 3. Lgwr process is active on the mount that contains the redo logs
>
> 4. Dbw0 process is active on the mount that contains the control and
> redo logs.
>
>
>
>
>
> Fuser –cu shows this activity.
>
>
>
> However, very rapidly the dwb0 process will appear in the mounts that
> contain only datafiles. This is messing me up as I am trying to unmount
> them. I have been running the fuser every 2 seconds to a log and capture
> the DWB0 process doing something on these mounts. The tool I am using
> expects these devices to be able to be unmounted.
>
>
>
Other related posts:
