1

I deploy my ignite cluster on google kubenetes engine (gke), which consists of two ignite server nodes.

I have a table with 6,000,000 records, and each around 220 bytes, in total 1.33GB. I set one backups for this table and set 5 indexes. As the ignite mentioned capacity planning here, these records will occupy 9.2G memory, and indeed in the graidgain control center I can see the total allocated memory is around 9.3G, but actually in gke console it shows about 18G.

I'm not sure what consumes this extra 9G memory resources.

Vincent Y
  • 109
  • 6
  • Hi, Could you please provide a metrics dump from the logs (can be found by 'Metrics for local node'). I guess that total allocated memory is the size of memory region (offheap memory), but as you understand JVM requires some memory for heap, so eventually it can be around 18G. – solveMe Mar 08 '23 at 09:20
  • yeah inded jvm heap takes some extra memory, thanks for your remind. – Vincent Y Mar 10 '23 at 15:44

1 Answers1

1

Here are some updates:

I tuned some JVM parameters including heap size, GC, and metaspace. In particular, I set 512m for max heap size, use G1 GC, 200m metaspace size and 40% min free ratio, 80% max free ratio of metaspace.

Now the memory consumption is reduced to 11GB, which is acceptable.

Vincent Y
  • 109
  • 6