High disk IO from ETCD
48 views
Skip to first unread message
Baek Stack
Jan 25, 2024, 5:13:36 AMJan 25
to etcd-dev
context: b/320538028#comment9
Hello, I am leading scalability effort for GDCV-BM (go/ekpbm) and we're seeing an issue where CP node(s) become unreachable (and un-sshable as well) when we're adding 500 nodes to the cluster. Though it is somewhat expected for clusters to be under high stress, I am investigating why the node would be stuck in unreachable state. I am currently suspecting high disk throughput from etcd caused kernel panic but I cannot find any details of such nature of high disk throughput. Any lead would be deeply appreciated. Thanks!
Hello, I am leading scalability effort for GDCV-BM (go/ekpbm) and we're seeing an issue where CP node(s) become unreachable (and un-sshable as well) when we're adding 500 nodes to the cluster. Though it is somewhat expected for clusters to be under high stress, I am investigating why the node would be stuck in unreachable state. I am currently suspecting high disk throughput from etcd caused kernel panic but I cannot find any details of such nature of high disk throughput. Any lead would be deeply appreciated. Thanks!
Reply all
Reply to author
Forward
0 new messages