We have a Percona XtraDB 5.6 cluster of 3 nodes. For some reason on two nodes, it takes an abnormal amount of time to run queries that involve tables in the information_schema database.
For ex, running “SHOW DATABASES” takes ~5min on node2 and ~3min on node3 whereas on node1, it takes ~0.02sec.
Node1 has a system memory of 8gb and innodb buffer pool size of 3gb.
Node2 and 3 have a system memory of 4gb and innodb buffer pool size of 2gb.
I know that the information_schema tables are stored in memory. I am theorizing that because of a small innodb buffer pool size, node2 and 3 take a long time to query against information_schema. However, when testing on a test cluster, I cannot reproduce this issue by lowering the innodb buffer pool size.
Has anyone ever experienced and know what the issue may be?