hello

I have a file system with 737 Go of data (oracle)
I want to add 230 Go.
IBM technician says to me that it's better (for performance) to backup the file system, rebuild it with the new 250Go and restore it....
737 Go to backup, it is not very simple... !!!!
You confirm what says the ibm tehnician ?

thank you for your help.

If you create a large filesystem and then load it with data, the data will be distributed more of less evenly across the entire filesystem. If you extend a filesystem, the free space is mostly at the end of the disk. Either one could be better than the other under various circumstances. Suppose you never add any data...the heads never will need to move to the last 25% (or so) of the disk. Suppose a file near the beginning of the disk grows... it will probably be extended into cylinder groups at the end of the disk. Now the heads must move from one end of the disk to the other end of the disk just to read that file. Because this is Oracle, you need an Oracle expert to comment on this.

But even if you simply extend the filesystem.... (Or even if you leave it at the same size!) you absolutely need a backup! What if you try to extend it and the process fails? If the size of this filesystem is so large that you are having trouble making backups, maybe you should not extend it. Maybe it is time for a new filestem instead. And talk to your DBA. You probably need to use some kind of Oracle backup. You cannot just backup the dbf files of a live database.

i agree. my impression would be that you would probably want to backup and rebuild for performance reasons. as for backups... you probably should already be doing that anyways.

If you cannot have your DBAs shutdown the database for the duration of your backup, perhaps they can put it into "hot backup mode"? This puts changes into separate files so that you can get a consistant backup of the 6 files oracle uses. You can't backup a "live" oracle database because if these 6 files aren't consistant, oracle won't touch them.

There is something about "PP size". I've yet to see this explained very well (I'm more of a Solaris guy). Essentially, you can only extend a filesystem so far. If you know you're going to expand a given filesystem, you can go in and mess with the numbers (when you build the filesystem) and make them big enough to expand things down the road (and what exactly this costs you; why you wouldn't just do this all the time, I don't have a clear understanding of either). If you just accept the defaults, then at some point, if you expand and expand and expand, you'll eventually reach a point where you can't expand anymore (something runs out of fingers and toes to count on). Whereas if you build a new filesystem from scratch, and you make it big, the defaults will bump up to some bigger numbers.

Sorry I don't have a better explaination; I don't understand it well myself. Perhaps someone who does understand all this can explain it better. I'm just thinking that perhaps this is what IBM is talking about.

Of course, you can call IBM back and ask them to explain it to you...

PS: It is my understand that a JFS2 filesystem doesn't have all this. I don't know when they added JFS2 or if that is an option for you.

with the original jfs, i think there was a limit of how many "PP"s you can have per physical volume... or something like that. with jfs2, i don't think there's limit that anyone would realistically reach for most applications. certainly, 1TB in this case could easily be attained.

thank you for your responses.
So i understand better: the new physical volumes should be at the end of the file system, and the heads must move from the beginning to the end.
Effectively the data are already saved, but there is always the risk of a corrupted backup... Or perhaps if I install 1To more, i move the file system to the new, i verify that all is right and I delete the old file system.
The file system is jfs2, and i have verified, i can extend it

hello

IBM advices to me that there is the possibility with reorgvg.
When i add a disk in a big volume group, the data is not written on the new disk, but Aix write on the old disks and after, when they are full, on the new disk.
With reorgvg, data are moved on each disk, so the new, for best performance.