Yep - reuse of a db instance, in serial, is fine - encouraged, even 🙂

Even if I would have multiple threads reading from it?

Reuse of a db snapshot isn't really applicable if you're writing as well - because it's a snapshot, it won't reflect later writes

Not from multiple threads at the same time, no - the queries on any one db instance should be in serial

If you want to query from multiple threads, it's best to take a db snapshot per thread

Somebody else may be writing but I want to make a decision as fast as I can. So I thought whether I could take a snapshot once and then parallelize the calculation. i.e. the write won't come to the paralellized db instance.

You can request that the snapshots from each thread be at the same 'basis', so that all the threads are seeing a consistent, immutable view of the database

Ah, that could work too.

I'd probably ask for xt/latest-completed-tx on the node, and then pass that to each thread to use in its db call

> the queries on any one `db` instance should be in serial clarification: does that mean you shouldn't use the same db to do other queries while iterating over the results of open-q ? I have code like that and it seems to work, is there some caveat

in the same thread still

> I have code like that and it seems to work, is there some caveat the only caveat I can think of is that completely unrelated queries will be clobbering each others' hot cache entries

thanks, it seemed obvious to use the same db handle that I opened using open-db as that docstring states it reserves shared resources to make multiple requests faster... haven't measured if there's an impact wrt hot cache vs opening multiple dbs

@U11SJ6Q0K the db instance (particularly ones from open-db) does reuse some state/resources between queries, as you say - so, to be on the safe side, I've usually taken out separate db instances. That said, there's been a lot of refactoring in that area since the last time I seriously analysed the implications, it may well be sufficiently safe now

just not something that we guarantee 🙂

perhaps there should be scarier warnings in the docstrings, I've had some crashes as I didn't understand the threading implications simply from the fn docs

Created this one instead: https://discuss.xtdb.com/t/old-kafka-config-and-should-i-change-it/74