Re: How do I get rid of the MAX_ROWS limitation?

[Benjamin J.]

Hi Mauritz, thank you for your answer.

Before 7.2.6 my workaround was to ADD PARTITION BY KEY() PARTITIONS 12;

I could not go higher than 12, because 16 (I have 4 datanodes) wasn't accepted, when I tried to alter the table. (I used ndbd, because I had SendBufferMemory problems with ndbmt.)
This workaround also made it hard to add datanodes.

When I don't partition manually, a hash is computed from the table's PK and assigned to a datanode using the modulo operation, resulting in a good distribution of my data over all data nodes. Adding datanodes leads to an automatic repartitioning of all data, right?

How does ALTER TABLE ... MAX_ROWS = 1000000000 exactly work? Can I still easily add datanodes without having to manually repartition?

And what is LDM? I have hyperthreaded quad core machines, but only two of them are at near 100% while the others are pretty lazy. Can you give a little more details on that please? Thank you!