I have a 5 node ES cluster running ES2.4.4. 3 nodes are physical servers and 2 of them are VMs. All nodes have the same number of cores allocated for ES and the same amount of memory. All nodes (ofcourse) run exactly the same version of Java and the same ES builds. I'm seeing that Marvel shows more frequent Garbage collection on the 2 VMs but less fequent on the physical servers. What would be the factors influencing this?
Thanks,
My company runs an all VM based ES cluster and as an experiment we reduced the number of cores available to half (from 64 to 32) on four of the nodes, the result was that garbage collection shot through the roof with gc overhead logged every few seconds on those four nodes. So you may want to check that your VMs have access to the same number of cores (htop) as the physical servers.
1 Like
This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.