DSXchange

I was wondering if anyone on HP Unix 11 using DS 6 Server has ever encountered a probem where the message is:

Error:ds_seqput: error in 'write()' - File too large


This occurs as the file crosses over the 2GB threshold. The job, when it runs successfully, will create a file 4GB's in size.

This is not a file system limitation as files can be created much larger than 2GB (currently the high water mark is 11GB).
This is not a space issue as when the job fails there is tons of DASD left.

This problem is driving us nuts since we cannot identify a cause.

Any ideas would be appreciated.

Andrew

Server jobs have a 32bit address limitation, so therefore you cannot write jobs larger than 2.2 gigabytes.

Second, you have the slowest design possible, because obviously you have a single job writing this file. This means you have a single cpu at work. You're on a unix box, so I hope you have multiple cpus. You should partition your source data such that you can have N instantiated jobs working on different subsets of the source data. You will cut your target file to 1/N in size, and probably squeeze under the 2.2 gigabyte limitation. You also will probably reduce your overall runtime by a factor of N. Your resulting files will also be easier to manage, and if you need to combine them a simple unix cat statement does the trick.

As I indicated in the original post, the same job will run successfully. So if it is a 32 bit addressing problem, then it is a random problem. We also have files (created by DataStage) of 11GBytes.

We are currently trying to get a 2CPU upgrade (we currently have 2) and then if that doesn't help a significantly larger Unix box may help our performance issues.

If there was any consistency in when it fails it would be great. However there isn't. It sometimes is a problem sometimes not.

However I will point out to the developer maybe they should try the split method.

Andrew

Wow, I missed that one. They used to have the limitation, I guess I need to get on the later versions. Sorry for the misinformation.

However, you need to think about how long a single job is spewing a file. If you're talking about a 4 hour runtime to produce the single file, multiple instantiations allows you to atleast break the work down to smaller units of work, which means that if 1 of 5 breaks, that's the only piece that needs to re-run. In your case, there's probably no resurrection capability, you simple start it over, re-incuring the total runtime. This can be frustrating, especially when it breaks at the end.

In addition, the instantiation approach positions you for more cpus. If with a 2 cpu box, you could potentially halve your runtime using 2 instances. Not knowing your job design, of course, will influence this capability. If you are querying a table and spewing the results, you could probably run a heck of a lot more instances if the database is remote, this is because the resources on the DS server are probably abundant.

Hi Stafford,
This is Error message is for the file system and you are trying to create a file on the file system which is more than 2 GB . Send a request to the unix admin to change the system from 32 bit to 64.
Hope this works
Tnks,
Neena