ManifoldCF's usage of memory is bounded per thread, but obviously scales with the number of worker threads you have. If you are using Tika, the amount of memory that may be used varies a lot, however, because Tika's streaming document memory behavior is quite variable, depending on the kind of document it is extracting from, and we have no control over that. So you need to set memory parameters experimentally when using Tika.
If you need to separate out who is causing memory problems and how much is needed, you can run Tika in "server" mode and use the Tika external transformer which will allow ManifoldCF to work with it. That's not as flexible but at least you can get a better idea of where the memory is going and who to complain to. Karl On Tue, Jul 24, 2018 at 5:43 AM msaunier <[email protected]> wrote: > Hello Karl, > > > > With my ManifoldCF, I have 12 Go used sometimes. > > I would like to know if certain actions make it possible to optimize this > use of memory. Commits more frequent for example? > > > > Thanks, > > Maxence, >
