I am testing out 7.1 on a two node cluster, running on a single server with 64G of RAM. By default, it looks like memsql-deploy assigned 56513 as the max_memory_mb value for both the LEAF and the MA. In the past I’ve always configured the LEAF and MA max_memory_mb to add up to 64G (or less). Can you explain the pros/cons of sticking with the default config?
No, its not a good configuration to have the sum of maximum_memory of both memsqld processes over 90% of the machines physical memory. You can end up with linux “out of memory” killing one of the memsqld processes if memory use gets too high (https://docs.memsql.com/v7.1/concepts/memory-management/#linux-out-of-memory-behavior). I would probably set your MA maximum_memory to ~10 GB and your leaf to ~47 GB or something along those lines. Ideally memsql-deploy would do this for you. I’m checking with folks internally on why its not adjusting maximum_memory automatically.