Cache: cannot write to disk, no space left

dataservices administration
1

Hi Guys (long time no see),

Since the weekend our production job server is throwing errors like this:

This happens sometimes, and sometimes not. And this happens for several repositories.

Also restarted the job server service.

I’ve checked:
[list]- Disk space (60 GB free)

  • Access to the folders and subfolders (BODS runs under local system)
  • Locked files (no files present)
  • Windows Event logs (only that al_engine did crash)
  • SAP Notes
  • Google
  • This forum[/list]
    And sadly enough found nothing. Does someone have a suggestion what the problem could be?

This is a similar topic:

Johannes Vink :netherlands: (BOB member since 2012-03-20)

2

My initial thought is that you may have been running parallel Jobs/Dataflows that both needed to write to cache. Run independently they would have worked fine. But together it was too much.

It could also have been other processes writing to D:.

I have jobs that write temp files as part of the processing. I had the admin create a separate area (drive) for the files to go to so that if I filled it up I wouldn’t cause problems with other jobs that were using normal caching, logs, etc.

eganjp :us: (BOB member since 2007-09-12)

3

True, multiple sys configs of this job are running.

I’ve dived in to the job and found several things that could be improved. Varying from better pushdown, stricter joins, better functional design etc. This brought down the load and the processing time a lot and now suddenly everything worked. Coming night it runs again in production and then we’ll for real.

I suspect it was a combination of a lot of things and in the end writing to the disk too fast with too many processes.

Unless with 2 instances of a job the same name for a cache is used? Does not seem likely, it looks like the PID is part of the file name.

Thanks for you thoughts.

Johannes Vink :netherlands: (BOB member since 2012-03-20)

4

If two jobs were trying to write to the same file I would expect a different error message.

eganjp :us: (BOB member since 2007-09-12)

5

True, but I would also expect another message than ‘Cannot write to disk, disk full’ when the disk not full :wink:

Anyway, last night the jobs ran very smooth. Still found some points for improvement, but that is for a later moment.

Johannes Vink :netherlands: (BOB member since 2012-03-20)