Why monitor the depot?

If you run an Eon Mode database on a cloud platform such as AWS, monitoring your depot in MC can help you tune performance and reduce expenses.

If you run an Eon Mode database on a cloud platform such as AWS, monitoring your depot in MC can help you tune performance and reduce expenses. MC can help address the following questions:

To access depot monitoring capabilities: from the MC home page, navigate to Database > Activity > Depot Activity Monitoring. See Monitoring depot activity in MC.

How often do queries hit the depot versus the S3 bucket?

Queries run faster when they access node-based depot data rather than fetch it from communal storage. For details, see Query Depot Hits and Misses

Is the depot optimally sized?

To optimize your queries for speed, you might want to resize the depot to fit your query workload. This ensures that queries do not need to spend extra time fetching data from the communal repository on S3. The Eon meta-function ALTER_LOCATION_SIZE lets you change depot size on one node, all nodes in a subcluster, or all nodes in the database. The following statement resizes all depots in the database to 80MB:

=> SELECT alter_location_size('depot', '','80%');
 alter_location_size
---------------------
 depotSize changed.
(1 row)

On the Depot Activity Monitoring screen, in the Communal Storage Access Calls chart, MC displays how many of each type of API call your queries executed in a given timespan. To see details on which queries were running, click on any point on the chart.

What is the current depot usage on each node?

The Depot Content tab of the Depot Activity Monitoring page provides detailed information about how each table is using the depot space on the cluster nodes.

Are projections and partitions tuned for best query performance?

On the Depot Content tab, when you select a row, you are selecting the table depot content on a node. MC loads the details for that table for that node in the bottom section of the page, to show depot content for the selected table, broken down by either projections or partitions on a given node.