mix raw chunks and cooked files possible in an emergency? 1

[WiccaChic]

Howdy all. I work on some ibm unix systems that use informix. We have a big main system and a smaller test system, and we want to save our database from our main system and restore it to test - just for some temporary testing. I am wondering a couple of things #1 how to compare the space available on both to make sure we can accomodate and #2 if we need some additional temporary space on the test system is it possible to use cooked files in addition to the raw space it alreadt uses? Dont have the disk on the test machine now for cooked files, but I do have plenty of flat, normal unix disk.

Performance is not an issue that would stop me from mixing cooked and raw if its possible. This is for functionality testing, so performance is not a concern.

 

[cdlvj]

Not real sure, but would think that to do this you would need different filesystems defined.

 

[vpshriyan]

Hi,

Use onstat -d on both servers which reports sizes in pages, that means figures need to be multiplied by 2 or 4 to convert it into Kb., according to your page size.

Mixing of raw and cooked file chunks for dbspaces are allowed. However, it is to be noted that if a dbspace already intiated with a raw chunk, one can not add cooked file chunk into it, and vice versa. If this this wrecks you plan, since you have ample flat, normal disk space, create them as much as needed. You need to resort dbexport from the main and dbimport for the test servers to implement this.

Regards,
Shriyan

 

[WiccaChic]

Thanks again vpshriyan, although this news is a bummer

I was hoping to mix different mediums for already existing file spaces in order to get my backup system looking like my production system from just the space/chunk standpoint.

 

[WiccaChic]

Hi vpshriyan! I hope you see this. I'm sorry but I think I misunderstood about your last post and not being able to mix the two. I went ahead and created a cooked file and added it to an existing dbspace that was totally made up of raw disk chunks and it worked.

So I thought I would ask you to clarify what you meant when you said "if a dbspace already intiated with a raw chunk." The new cooked-chunk added in okay and no errors in the log and I had expected it to fail.

 

[vpshriyan]

Hi Wic,

One would experience better performance and increased integrity / reliability when one use raw disk space. A raw disk space, when bound by a character-special interface, leaves the buffering management to an application. In contrast a cooked file system always attract buffering at OS level. If we mix these two in one dbspace, as I wonder that, is an Informix dbserver is equipped with mechanism to sort this out accordingly? I have never tried to mix the chunks in this way. We have combinations of Raw and Cooked disk space been used at our end. I never tried these "jinx", which contradicts fundamental behaviours; which *may* result into future unavailability of database services to it's users.

Regards,
Shriyan

 

[WiccaChic]

I contacted support and was told that doing this was possible, but not recommended. I am guessing that its not recomended due to the performance and reliability issues you mentioned, but I am just trying to get my dbspaces synched up in order to allow an Ontape restore between two systems after which I can clean out some tables and drop the cooked space in a pinch - so I think in my case the risk versus reward are okay.

Thanks for your expertise.

 

[WiccaChic]

Hey all. I did this mixed chunk thing and it actually worked out okay. I did have one weird thing happen the first time I fired IDS up, the chunks got marked down immediatly and I about wet my pants. But I did an &quot;onspaces -s <dbs> -p <path> -o <offset> -O&quot;
on each one and they all came right up.

Not sure what happened and I would certainly think twice before doing this in a mission critical environment, but it did allow me to at least get these two systems in synch and restored so hoora!

Thanks again.