Re: Row Devices and Chunk offset
Posted in 1997
Topics: Installation, Setup & Upgrades, Storage & Space Management, Platform-Specific Issues
The requirement for an offset is documented in the man pages in AIX 4.2 under the make logical volume command ("man mklv"). Under number 2 in the Notes section, it tells of a 512-byte control block which should not be overwritten. If the block is overwritten, things will probably run fine until you someday try to do something with the volume group like add another volume, rearrange the logical volumes, etc. Of course, it will break at the worst possible time. :) Since the Informix page size for AIX is 4KB, this should be the minimum offset in a raw logical volume, since I believe Informix wants chunks on page size boundaries. -- Ken Guenther ---------------------------------------------------------- kguenthe@netacc.net -- The usual disclaimers apply Axiom Corporation, Rochester, NY, USA (716)654-1100 ---------------------------------------------------------- Dragi Raos wrote: > > sujata_soman_at_omm-la2-infotech-001@internet.omm.com wrote: > > > > Hi all. > > We are running Informix 7.22 on out IBM machines ( AIX ) and have > > not specified any offset. I do not know why you are facing any > > problems. There is no such requirement that I know of > > Can you give me some more specifications ( since I deleted your > > original message by mistake ) and then I may be able to help you. > > > > Sujata > > > > ______________________________ Reply Separator _________________________________ > > Subject: Re: Row Devices and Chunk offset > > Author: Jonathan Leffler <johnl@informix.com> at internet > > Date: 10/15/97 8:54 AM > > > > On Wed, 15 Oct 1997, Philippe Meylan wrote: > > > > > We are using INFORMIX 7.1x and our chunks are located on row devices. IBM > > > inform us that an offset of 512 bytes is required between the base of the > > > row device and the beginning of the chunk. This area should be used by AIX > > > for system management. > > > > Not an unusual requirement, but a confounded nuisance. Many o/s make that > > sort of requirement, but often for more than 512 bytes. > > > > > My question : > > > > > > This requirement is undocumented > > > > By IBM? You seem to have found out about it... > > > > By Informix? It varies from platform to platform -- many machines don't > > need it; some do. On some machines, it depends on which raw device you are > > using. And, in any case, p15-6 of the version 7.2 OnLine DSA Server > > Admin's Guide discusses 'do you need to specify an offset?'. > > > > > and the ONINIT tools doesn't reserve any offset when creating a new > > > chunk. > > > > Unless you tell it to. One of the main purposes of the optional offset > > parameter is to avoid treading on the information required by the system. > > (The other purpose of the offset is to allow multiple chunks to share a > > single raw disk partition.) > > > > > It's this requirement justified > > > > This requirement is justified on machines where it is required, and not > > on machines where it isn't. > > > > > and can be applied on any platform? > > > > It can be used on any platform, but will waste space on platforms where it > > isn't needed. > > > > > Has Informix a document explaining that ? > > > > See the OnLine Admins Guide. > > > > Yours, > > Jonathan Leffler (johnl@informix.com) #include <witticism.h> > > > > > > Well, I have seen AIX and Solaris installations (both 5 and 7) without > offset that work fine. On the other hand, when you create a fresh > logical volume on AIX and dump first several bytes, they do contain > something (LV name etc). Curious. > > Dragi "Bonzi" Raos
Related threads
- Using a "+" character in an SQL insert statement?
- Re: Delay in commands returning
- Set explain what does it mean?
- RE: Using a "+" character in an SQL insert statement?