DSXchange

Search also for bad file(s), possibly on the scratch disk for the partition indicated in any error message.

Are you trying to read from the File Set using a different configuration file (APT_CONFIG_FILE) than the one that was being used when the File Set was created?
The fact that there are node names not found suggests to me that that's the root cause of the problem that you are encountering.

There is no bulk load facility of which I'm aware. ODBC is best because (again, as far as I am aware) it is the only mechanism for accessing MySQL from DataStage. Or any other client for that matter.

Please post your metadata. We can not tell without knowing that. The record schema will suffice.

What version are you using, and what FixPack levels have you already applied?

It can, but only if one of the uvconfig parameters is changed to permit it.

The error on sh is probably because your PATH is not set up properly to include the MKS Toolkit bin folder.

Welcome aboard. Do you have many of these substitutions to perform? If not, how about an input text override in the rule set?

It's probably easiest to reformat using a DataStage job.

This will stand you in good stead for the future, when QualityStage is much more tightly integrated with DataStage, and operators such as FFC disappear from QualityStage.

Because there is a minimum quantum of data that DataStage moves at a time. 24/26 is close enough to evenly spread.

For what it's worth, Round Robin is almost certainly completely wrong for your Join and Lookup stages.

Why? Memory is a resource to be used. Is there spare memory available? Then don't worry about it. You have not told us much about these jobs. Do they use hashed file cache, do they do sorting, are there any Aggregator stages? These are just some of the components that will make larger demands on mem...

The philosophy is "one stage, one task". If ALL you want to do is copy, then a Copy stage is way more efficient. The Transformer stage is an aberration. It's been put in so that you don't have to create buildops for every single transformation. These days, if you examine the code generated by compil...

The workaround is obviously to do it right. How many outputs do you have? What do you think transferAndWriteRecord(1) will achieve?

Any solution for this? Doubt it. You are trying to make AT do things that it just doesn't do. Or, at least, wishing that you could. I guess, if you knew enough about Windows internals, you could write some code that interrogated the internal structures where AT stores its queued jobs, but that's ou...

If you do not enable pre-load to memory each record is logically read from disk. In practice very few of these I/Os will be physical I/Os, because the relevant pages are likely to be in memory - either main memory or in the disk controller's memory. Even so a hashed file is likely to out-perform eve...