GitHub user nickaccuweb added a comment to the discussion: High Latency Between Two VMs In a L2 Private Guest Network
@fnavidan Good to hear that you found the cause of the issue and that GlusterFS performance has improved after stopping the crypto miner. Regarding your question, I think 10 Gbps should be sufficient to start with for your GlusterFS setup with 2 replicas and 1 arbiter. However, it also depends on the amount of data being written and the number of VMs accessing the storage. Since you are planning to move each GlusterFS VM to a separate KVM host, I would suggest using a dedicated 10 Gbps network for GlusterFS traffic, if possible. This will help avoid network congestion from other VMs. Once you move the VMs, run an iperf3 test between the nodes to check the actual available bandwidth. You can also test GlusterFS write performance again to see how much difference the physical network makes. One more thing worth checking is your NFS primary storage. If all three GlusterFS VMs are using the same underlying NFS server, that storage will still be a single point of failure, even if the VMs are running on separate KVM hosts. For reference, you can check the official GlusterFS documentation: https://docs.gluster.org/en/latest/Administrator-Guide/Network-Configurations-Techniques/ Here, I would start with 10 Gbps and monitor the performance before considering a network upgrade. GitHub link: https://github.com/apache/cloudstack/discussions/14380#discussioncomment-19017939 ---- This is an automatically sent email for [email protected]. To unsubscribe, please send an email to: [email protected]
