Wednesday, April 17, 2024

DBMT is good - but like most hybrids, it is a compromise

Sometimes I think of DBMT as the love child of Compact and Updall. It combines some of their functionality, but sometimes you notice bits are missing, don’t work, or don’t work the way you think they should.

The big issue that we keep encountering lately is that DBMT only enforces a time limit for it's compacting tasks; the index update threads will happily continue for hours. This is not ideal, especially for some view indexing operations on large databases.

I have discussed this before, and now I'm leaning more towards reverting to the old practice of running Updall as a server task and simply setting the updall threads in DBMT to zero. After all, it worked perfectly fine like that for over 25 years!

It feels like the right way forward. 

Can you see any obvious downsides? I’d love to hear your thoughts on this.

No comments: