Unix Administrator stikes again

The Unix administator has just ruined my Saturday (familiar story ?). We were replacing 4GB disks with 18GB, I marked the chunks on the disks to be removed as DOWN, relying on the mirror to recover later. But one wrong disk was removed and now I have 2 chunks on the same dbspace with both primary and mirror (which failed sanity checks) marked as DOWN. I know Informix can log on and "fix" the chunk state, but this is the weekend .. I intend running "onbar -r dbspaceX" to restore just the effected dbspace. The manual states that you can not restore an indiviual dbspace to a point in time, so my question is what of data consistency ? There was very little if no database activity since the last Level 0 but I can not be sure. Is a full level 0 my only "safe" action ? Hardware: Sun E450 x 4CPUs OpSys: Solaris 2.6 (SunOS 5.6) Informix Products: Universal Server 9.14.UC2 + WebBlade 3.31.UC2 ------------------ Geoff Poole Informix Database Administrator geoffrey.poole@ogs.co.uk Oxford GlycoSciences (U.K) Ltd,

>The Unix administator has just ruined my Saturday (familiar story ?). >We were replacing 4GB disks with 18GB I hate to be wise "after the event", but it's difficult to imagine that you would undertake a major piece of work like disk movement without taking a level 0 first? > There was very little if no database activity since the last Level 0 Maybe you'll be OK with your onbar restore then. Geoffrey Poole wrote in message <86uk0h$fle$1@news.xmission.com>... > >The Unix administator has just ruined my Saturday (familiar story ?). >We were replacing 4GB disks with 18GB, I marked the chunks on the disks to >be removed as DOWN, relying on the mirror to recover later. But one wrong >disk was removed and now I have 2 chunks on the same dbspace with both >primary and mirror (which failed sanity checks) marked as DOWN. >I know Informix can log on and "fix" the chunk state, but this is the >weekend .. > >I intend running "onbar -r dbspaceX" to restore just the effected dbspace. >The manual states that you can not restore an indiviual dbspace to a point >in time, so my question is what of data consistency ? There was very little >if no database activity since the last Level 0 but I can not be sure. Is a >full level 0 my only "safe" action ? > >Hardware: Sun E450 x 4CPUs >OpSys: Solaris 2.6 (SunOS 5.6) >Informix Products: Universal Server 9.14.UC2 + WebBlade 3.31.UC2 >------------------ >Geoff Poole >Informix Database Administrator >geoffrey.poole@ogs.co.uk >Oxford GlycoSciences (U.K) Ltd,

In article <86uk0h$fle$1@news.xmission.com>, Geoffrey Poole <Geoffrey.Poole@ogs.co.uk> wrote: > > The Unix administator has just ruined my Saturday (familiar story ?). > We were replacing 4GB disks with 18GB, I marked the chunks on the disks to > be removed as DOWN, relying on the mirror to recover later. But one wrong > disk was removed and now I have 2 chunks on the same dbspace with both > primary and mirror (which failed sanity checks) marked as DOWN. > I know Informix can log on and "fix" the chunk state, but this is the > weekend .. > > I intend running "onbar -r dbspaceX" to restore just the effected dbspace. > The manual states that you can not restore an indiviual dbspace to a point > in time, so my question is what of data consistency ? There was very little > if no database activity since the last Level 0 but I can not be sure. Is a > full level 0 my only "safe" action ? > > Hardware: Sun E450 x 4CPUs > OpSys: Solaris 2.6 (SunOS 5.6) > Informix Products: Universal Server 9.14.UC2 + WebBlade 3.31.UC2 > ------------------ > Geoff Poole > Informix Database Administrator > geoffrey.poole@ogs.co.uk > Oxford GlycoSciences (U.K) Ltd, > Geoff, My experience has been that if you restore a single dbspace, in addition to the steps listed in the manuals, you have to have an Informix engineer mark your chunk consistent. This has to be done for exactly the reason that you mention; the database itself cannot guarantee that there hasn't been activity in the database that would invalidate your restore. You'll need to provide Informix an authorization document to do this as well. Not every engineer requires it (some will accept that there is a document on file), but most do. It states (in general terms) that you accept responsibility for the data in the database, and acknowledge that marking the chunk consistent is a risky thing to do, if you don't know your data (and you already know that). If you, or anyone else has had different experience restoring a single dbspace, please let me know. -- Dan Michaelis Database Administrator dan@kax.com Sent via Deja.com http://www.deja.com/ Before you buy.