| I think that most statistics we have are useful, especially the ones about 2nd-level cache hit/miss count. I would add the following metrics:
- query response time histogram (min, max, avg, percentiles)
- number of SQL statements being executed (for SELECT, INSERT, UPDATE, DELETE) per transaction, so that we can figure out if batching makes sense, or there's some N+1 issue
- transaction response time histogram (min, max, avg, percentiles)
- transaction request count (how many incoming transactions we have)
- transaction count (how many transactions we usually execute at once - this will be correlated with the connection pool size probably)
|