The recommendation for heap size of the historical seems confusing in the document


Thanks for all the effort you’ve done for the community.

I’m building a new cluster now and I’m searching for recommendations about the heap size for the historical.

However, in the document, the recommended heap size is a little bit confusing.

In the performance FAQ:

It says 250MB * numThreads.

And in the production cluster configuration:

It says 1GB per core.

However, in the historical configuration document:

It says numThreads should be core number - 1, which seems to make the two recommendations above self-contradictory.

For instance, I have an 8-core, 32GB-memory server for one historical node.

According to the document, I should set the numThread to 7.

Then, for heap size, should I set it 250MB * 7 or 1GB * 7?

Or maybe I misunderstood something here?

Thanks in advance!