hello, we have an issue at work.
we kick off reorg on multiple objects in a script. each object is reorged in a loop.
worker processes and max parallel degree are both set to 20 dynamically at the start of the reorg (and set back to default at the end of the script).
but somewhere between reorg of one object and another, some other transaction comes in and hijacks the worker processes. then reorg is left to run single thread and takes a long time to run.
questions :
1. like the consumers option with create index or update stats, is there a similar option for reorg?
2. will consumers option work on non-raid disks? one of my collegues told me (I think) that because the disks were on san, Sybase couldn't know if they were not raid or not (or not striped), or something to that effect. and so will consumers option work even on non raid disks? any opinions?
3. is there any other way to bind parallelism to a particular transaction - like reorg for instance - and not let other processes hijack the worker processes meant for that function?
appreciate the insights.