On the Aura Console metrics tab, the last metric is disk usage as a percentage of total disk space.
How do you find out the actual size of the datastore on disk? The size represented in the chart includes any fragmentation caused by datastore optimization (id-reuse) as described in this article. It is important to note that the metric displayed in the console is the average over the three disks in your cluster (leader and two followers). If you ever need a good and exact value take a snapshot and load it to a fresh database. This will give you the defragmented data store size.
There is some margin above and beyond what is visible on the metrics page so even if you are showing 100% you have a bit more overhead than that.
- We actively monitor datastore and would proactively notify if you were getting close to the actual physical value.
- Snapshots are compressed. A ratio of 3-5:1 when expanded is very common but varies.
- Fragmentation is a natural disk optimization and only downtime and copy of the store (neo4j-admin copy) can address this.
- For disk space requirement estimation based on Rel and nodes, you can use: https://neo4j.com/hardware-sizing/. This provides a rough approximation of how large your instance should be.
- Maintaining control over the number of nodes and number of relationships is a good base metric but it does not take into account indexes, transaction logs, etc.
- This query will count your nodes `MATCH (n) RETURN count(n) as count`.
- This query will count relationships `MATCH ()-[r]->() RETURN count(r) as count`
Also you can use https://neo4j.com/labs/apoc/4.2/overview/apoc.meta/apoc.meta.stats/ to get the same information.