That means it needs some space to clear the dbi. One may transact some more things to it, so the map (file) is enlarged, then do the clear.

didn’t think this will happen, interesting

care to tell more details?

anyway, 0.9.0 will have fix, we will enlarge the map automatically if this happens.

You mean there’s an error? Cannot reproduce an error with your code above.

as to using db instead of conn, sure, it was not my first choice, if people feel it is more natural to clear a conn, we can do that.

0.9 will have this change

that is basically how I was setting things up - after that its basically shoveling data in.

which platform are you on? it seems to be some other issues.

“environment mapsize reached” is not something i see normally. During transaction, auto resizing will happen. not when clear. Since you seem to see it a lot, you need to tell more about your environment.

I'm running on arch linux, openJDK 19.0.2

I seem to see that spectre mostly after having loaded a singificant number of entiites. I continue to keep chugging through the ETL. Currently have roughly 12M datoms loaded, but that number creeps up as I progress

it would be helpful to know what function call trigger this.

d/clear

right now, we catch an exception when the map size is reached, then enlarge the map. this only happens in transact right now, so if other functions may trigger this, we can catch that and do the same

can you transact something before calling that?

the transaction will trigger resizing, if that’s the reason, but it may be something else.

currently in my repl.

at this point, some number of transactions have completed, and I'm just hanging onto the cn

you can try to transact something

after the prior clear attempt, the cn was closed. Reopened it here, transacted something, and retried the clear

I haven't noticed that it cares if I keep transacting even large numbers of things. Or retracting for that matter. The clear gets in trouble though

file an issue, and I will investigate

maybe it’s something about dropping the dbi in LMDB

I can, anything I can provide that will help with this? I can't give you the actual data, it's client proprietary.

no need

maybe I can sythesize the behavior with fake data.

you can set a tiny mapsize and replicate

e.g 1mb

It is affected by datom head count, though.

and create some fake data easily

I'll play with it and see where it starts breaking down

don’t think it is about datalog, it’s lmdb drop dbi logic issue

sure

maybe a bug in lmdb

#196, trivial to repro

thx

0.8.9 should have a fix. please let me know if it works.