Hi all,

This week we also provide a recommendation for RDS instances used by data team. This will be the last RDS recommendation from backend-infra team, because in July some RIs will expire. But please don't stop improving your resource utilization. We need your help:

Please follow our recommendation to change your RDS instance type (based on data from 2020-06-26): https://docs.google.com/spreadsheets/d/1Sv_XXIDDrLNEgZr3tFm9NlfhNYQp4Xe4r2Yn8UOGVyo/edit#gid=0 -> sorted by the most expensive instances first

Full instance list is at: https://docs.google.com/spreadsheets/d/1GxvPcOnoRQiyp8tvZwsHcIWNy16W43qQOjYTMJcuI74/edit#gid=0
Please closely monitor your metrics (especially latency and disk IOPS) after you change the instances or decommission replicas
Please update/comment the status in the "Status" column -> if you already tried our recommendation and it doesn't work, please state it in the "Status" column too
Using m3 or r3 is a waste of money
Using instances that are too big for our usages (CPU usage < 50% and memory usage < 50%) is also a way of wasting money
For t* families, we still overuse our RI, so please use smaller instances if your utilization is too low (related to the point above):

purchased RI units: 4824
used units: 7694

purchased RI units: 1188
used units: 2979

Please read and follow the recommendation for RDS replica: https://docs.google.com/spreadsheets/d/1n0QzuCZpDt_9rmBlsKjSQ8lT0I-n8YK2bSJPc2BF6Rs/edit#gid=259699164
Please enable monitoring for Hikari connection pool: https://app.slack.com/docs/T02T3CAFM/F012NEWMG2F?origin_team=T02T3CAFM

It is needed to tune your number of connections -> better utilization and performance of RDS

Please inform us if you have other plans for your RDS instances

20200629 - Change RDS Instance Type to Reduce Cost