Tuning the filesystem offloader performance

[Ultranium]

During data migration to Pulsar, I found that the filesystem offloader does not keep up with incoming data.
The offload threshold is set to 10GB, yet I can see not offloaded ledgers way beyond this mark (100+GB ago).
If I stop incoming messages and just wait, eventually it manages to offload all the data, so the problem is not a topic misconfiguration, but the performance of the offloader.

I use local storage as offload target, and I can see that its bandwidth is definitely not saturated.
Also, when there is no other load on the broker (no active producers or consumers), it uses only about 1 CPU core.
I looks like the offload processes just a single ledger at a time, and does it using a single thread.
Looking at the broker metrics, there is no intensive JVM garbage collection.

My question is, is it possible to somehow speed up offloading data? Maybe make it use more threads or something like that.
I didn't find anything related in docs.

I use a standalone Pulsar installation with a single broker.

[dave2wave]

I don't recall the name of the configuration, but the default number of off-load threads is 2 and you should bump that up to at least 8.

[Ultranium]

Thank you, the actual config option that controls offload threads is managedLedgerOffloadMaxThreads.

For some reason, I thought that all the offload-related settings listed here, while I should've also checked the Broker Configuration Reference