Attaching Informix .dat and .idx files

问题:

We are trying to duplicate one of our informix database on a test server, but without Informix expertise in house we can only guess what we need to do. I am learning this stuff on the fly myself and nowhere near the expertise level needed to operate Informix efficiently or even inefficiently. Anyhow... We managed to copy the .dat and .idx files from the live server somewhere. Installed Linux and the latest Informix Dynamic Server on it and have it up and running.

Now what should we do with the .dat and idx files from the live server? Do we copy it somewhere and it will recognize it automatically?

Or is there an equivalent way like you can do attach DB from MS SQLServer to register the database files in the new database?

At my rope end...

回答1:

You've asked a pretty complicated question without realizing it. Informix is architected as a shared everything database engine, meaning all resources available to the instance are available to every database in that instance. This means that more than one database can store data in any given dbspace, .dat or .idx file in your case. Most DBA's know better than to do that but it's something to be aware of. Given this knowledge you now know that the .dat and .idx files do not belong to a database but belong the instance. The dbspaces and files were created to contain your databases data but they technically belong to the instance. It's worth noting that the .dat and .idx files are known to the database by the logical dbspace name.

Armed with this background info and assuming that the production and development servers are running the same OS and that your hardware is relatively the same, not a combination of PARISC, Itanium or x86/x64, I'll throw a couple of options out for you.

1. Create the dbspaces that you need in the new instance and use onunload and onload to copy the database from production to development.
2. Use ontape or onbar to backup the entire production instance and restore it over your development instance.

Option 1 requires that you know what the dbspaces are named and how large they are. Use onstat -d on the production instance to find this out. BTW, the numbers listed in onstat -d are in pages, I believe that Linux is a 2K page.

Option 2 simply requires that the paths for the data files are the same on both servers. This means that the ROOTDBS needs to be the same in both instances. That can be found by executing onstat -c | grep ROOTDBS

There's a lot that has been left out but I hope that this gives you the info that you need to move forward with your task.

回答2:

The .dat and .idx files are associated with C-ISAM, or, when organized in a directory called dbase.dbs (where dbase is the name of your database), the .dat and .idx files are associated with Informix Standard Engine, aka Informix SE. SE uses C-ISAM to manage its storage. SE is rather different from (and much simpler than) Informix Dynamic Server (IDS). It is not impossible that the .dat and .idx files are associated with IDS; it is just extremely unlikely.

From the information available, it sounds as though your production server is running SE. To get the data from SE to IDS, you will probably want to use DB-Export at the SE end and DB-Import at the Linux/IDS end. Certainly, that is the simplest way to do it.

There are other possible solutions - C-ISAM datablade being one such - but they are more expensive and probably not warranted. There are other possible loading solutions, such as HPL (High-Performance Loader).

For more information about Informix, either use the various web sites already referenced (http://www.informix.com is a link to the Informix section of IBM's web site), or use the International Informix User Group (IIUG) web site. There are mailing lists available (which require you to belong, but membership is free) for discussing Informix in detail.

回答3:

Those Informix-SE datafiles (.DAT) and their associated index files (.IDX) are useless unless you also have all the associated catalog files, such as SYSTABLES.DAT SYSTABLES.IDX, SYSCOLUMNS, SYSINDEXES, etc.

Then you also have to worry about which version of Informix-SE created them, as some have a 2K or 4K index file node size.

Your best approach is to obtain all the .DAT and .IDX files from the source db, plus the correct standard engine, installed on the same hardware and operating system it came from.

Long story short, on the source machine, run "dbexport" to unload all the data to ascii files, and run "dbschema" to generate all the table schemas and indexes. It also wouldn't hurt to run a "bcheck" on all the files before unloading them to ascii flat files.

回答4:

I don't have any Informix-specific advice but for situations like this you can usually find the answer by looking up how to move a database (a common admin task, and usually well described in the manual) and just skipping the steps that would remove the old database.

Also, be careful of problems caused by different system architectures; some DBs fail spectacularly if you move them from a big-endian system (such as Solaris) to a little-endian system (such as x86 Linux) Again, the manual section on moving a DB would cover any extra steps that are needed.