Re: RMAN 'delete obsolete' command takes ages
From: Steve Howard <stevedhoward_at_gmail.com>
Date: Mon, 12 Jan 2009 07:03:09 -0800 (PST)
Message-ID: <d382edf5-78a8-45d4-bd04-2cfc0337d73e_at_r15g2000prd.googlegroups.com>
On Jan 12, 9:23 am, Ian Chard <ian.ch..._at_sers.ox.ac.uk> wrote:
> On 12/01/09 12:49, Mladen Gogala wrote:
>
>
>
> > On Mon, 12 Jan 2009 09:16:35 +0000, Ian Chard wrote:
>
> >> On 09/01/09 17:25, Steve Howard wrote:
> >>> On Jan 9, 10:25 am, Ian Chard <ian.ch..._at_sers.ox.ac.uk> wrote:
> >>>> Hi,
>
> >>>> I'm running Oracle 10g on RHEL 5.2, backing up with RMAN to Tivoli
> >>>> Storage Manager. When I run 'delete obsolete', it quickly lists what
> >>>> it wants to delete but then just sits there for many tens of minutes,
> >>>> occasionally deleting one.
> >>> select inst_id,event,seconds_in_wait from gv$session where program like
> >>> 'rman%';
>
> >>> What does that return?
> >> After waiting for a few minutes, that query gives:
>
> >> 1 SQL*Net message from client 201
> >> 1 SQL*Net message from client 204
> >> 1 Backup: sbtinfo2 198
> >> 1 SQL*Net message from client 201
>
> >> After ten minutes the same four entries are returned, but with larger
> >> seconds_in_wait values.
>
> >> Does this tell us anything useful?
>
> > Yes it does. RMAN is waiting for a reply from your backup software and
> > is not getting it. The MML layer is not properly configured. What backup
> > software are you using? NetBackup? NetWorker? Tivoli? Oracle Secure
> > Backup?
>
> It's Tivoli. Backups otherwise work correctly (I've done a full test
> restore which was successful).
>
> Another poster suggested that RMAN might be trying to read a small
> amount of data from each backup piece before deleting it. Does that
> sound likely?
>
> - Ian
>
> --
> Ian Chard, Senior Unix and Network Gorilla | E: ian.ch..._at_sers.ox.ac.uk
> Systems and Electronic Resources Service | T: 80587 / (01865) 280587
> Oxford University Library Services | F: (01865) 242287
Date: Mon, 12 Jan 2009 07:03:09 -0800 (PST)
Message-ID: <d382edf5-78a8-45d4-bd04-2cfc0337d73e_at_r15g2000prd.googlegroups.com>
On Jan 12, 9:23 am, Ian Chard <ian.ch..._at_sers.ox.ac.uk> wrote:
> On 12/01/09 12:49, Mladen Gogala wrote:
>
>
>
> > On Mon, 12 Jan 2009 09:16:35 +0000, Ian Chard wrote:
>
> >> On 09/01/09 17:25, Steve Howard wrote:
> >>> On Jan 9, 10:25 am, Ian Chard <ian.ch..._at_sers.ox.ac.uk> wrote:
> >>>> Hi,
>
> >>>> I'm running Oracle 10g on RHEL 5.2, backing up with RMAN to Tivoli
> >>>> Storage Manager. When I run 'delete obsolete', it quickly lists what
> >>>> it wants to delete but then just sits there for many tens of minutes,
> >>>> occasionally deleting one.
> >>> select inst_id,event,seconds_in_wait from gv$session where program like
> >>> 'rman%';
>
> >>> What does that return?
> >> After waiting for a few minutes, that query gives:
>
> >> 1 SQL*Net message from client 201
> >> 1 SQL*Net message from client 204
> >> 1 Backup: sbtinfo2 198
> >> 1 SQL*Net message from client 201
>
> >> After ten minutes the same four entries are returned, but with larger
> >> seconds_in_wait values.
>
> >> Does this tell us anything useful?
>
> > Yes it does. RMAN is waiting for a reply from your backup software and
> > is not getting it. The MML layer is not properly configured. What backup
> > software are you using? NetBackup? NetWorker? Tivoli? Oracle Secure
> > Backup?
>
> It's Tivoli. Backups otherwise work correctly (I've done a full test
> restore which was successful).
>
> Another poster suggested that RMAN might be trying to read a small
> amount of data from each backup piece before deleting it. Does that
> sound likely?
>
> - Ian
>
> --
> Ian Chard, Senior Unix and Network Gorilla | E: ian.ch..._at_sers.ox.ac.uk
> Systems and Electronic Resources Service | T: 80587 / (01865) 280587
> Oxford University Library Services | F: (01865) 242287
There is at least one bug on Metalink for 10G (but it's on AIX) for this type of thing. Its text says something to the effect of "NetBackup is failing to send some trigger messages to RMAN". Admittedly, it's not Tivoli, but it shows there can be issues with the storage vendor API with Oracle.
I would suggest you open an SR with Oracle, as they will probably be able to quickly identify if it is a vendor config issue. Received on Mon Jan 12 2009 - 09:03:09 CST