Hash file Location

[yiminghu]

Hi,

I'm having a question regarding to hash file. When you create hash file, you can use account name (usually I select this choice, and leave the account name empty) or use direct path.

We have a datastage consultant doing consulting service in our company.He suggested us we use direct path instead of account name. His reason is that by using direct path, we can bypass uvtemp directory, (because sometimes we have problem about 'can not open file' in the folder). Is it true? Besides that, any other reasons?

I really appreicate if somebody can help clarify this.

Thanks,

Carol

[ArndW]

Hello Carol,

It is usually a very good idea to separate your program (i.e. DataStage Project Account) and data storage. It is so easy to run out of data space by getting larger than expected data volumes, forgetting to delete large temporary files or many other possible reasons.

Normally this is not a huge issue; but DataStage is particularly finicky when it comes to a disk-full condition while writing within the project (i.e. saving a job, compiling a job, or writing to the log file while running a job). When this happens a project may become corrupt and require some work to get back into a runnable state - just do a search in this forum on that subject.

The uvtemp should be specified and be located elsewhere, so that is not directly the issue, but I think that your consultant has it right. Just make sure that the path is on a different volume, not just a different directory on the same volume as your projects and/or server.

[yiminghu]

Can I put paramter in this direct path? If not, that means the path is hard-coded in job, if I move projects from development server to production server, the same path has to be on the production server.

Carol

[kcbland]

You might find this information helpful regarding hash files:
viewtopic.php?t=85364

[ArndW]

By all means use a parameter for the path (usually the filename is fixed). I most often used "HashFilePath" as the parameter name [nb - it is best to have the path not contain the final delimiter] and then specify the path as #HashFilePath#/MyHashFile.

[ArndW]

Ken - that wasn't a thread on hash files, it was a BOOK! And the thread should have been titled "everything you wanted to know about Hash Files but were afraid to ask". Good job!

[yiminghu]

Thank you for all the info.

I still have one more question, if I use external directory path, how I do check the content of hash file using sql statement? ( I usually exeute through datastage administrator - command button). I found this feature is very convienient during the development of job.

Carol

[kcbland]

Search the forum for discussions on SETFILE.

[chulett]

I think I read somewhere that you can actually use SETFILE to create a VOC record for the parent directory that your hash files live in. This would allow you to treat the directory as an 'account' and so avoid having to add individual VOC records for each hash file in the directory.

Is this true? Can something like this be done?

[ray.wurlod]

Such a pointer can be created, but it does not treat the directory as an account. Rather it treats the directory as a table, the file names within the directory as key values, and the file contents as data records. Not all of which will be readable (hashed files, for example) via queries.

Also, you probably then need to modify the VOC entry to use a dictionary of directories. For example:

Code: Select all

UPDATE VOC SET F3 = '<<F(VOC,&UFD&,3)>>' WHERE @ID = 'MyPointer';

But this is not a solution for Carol's question.

[chulett]

I remembered where I found the little tidbit I was referring to - here. Curious what anyone thinks of the specifics of this 'tip'.

I haven't tried to implement it, so can't vouch for it... just had it filed away in the back of my brain.

[ArndW]

Craig,

I looked at the code, and it makes sense and ought to work. A bit of a roundabout way to edit the UV.ACCOUNT record but it does the job. It should be noted that this is not a valid DS account that one can attach to from the client front-end, but if you do specify this account name in the hash file stage it will work.

p.s. I did notice that my UV.ACCOUNT file at 7.5 is read-only.

[chucksmith]

I have been able to update UV.ACCOUNT on a 7.5 windows-based server. I was logged into the server under the user id that installed DataStage.