Compose Notes are designed to keep you up-to-date on the smaller changes occurring within Compose. In this edition, we're looking how we've enhanced the display of our deployment metrics.
Compose's metrics for capsules in deployments has just had a major rework. Gone are the semi-static views of old and in comes a new dynamically updated view with a lot more information. Here's a sample:
The purple memory limit line is the hard memory limit for the deployment while the light blue areas is the actual memory use on the node. We don't autoscale against that hard memory limit. There's a blue line, the memory soft limit, which is the scaling configured limit. On a memory based database like Redis, going beyond this soft limit would make the database a candidate for autoscaling.
An orange/brown line on the graph shows the percentage of swap in use on the capsule. A red bar graph shows where there were failures to immediately allocate memory - these may just be a reflection of the database putting pressure on memory to get performance or respond to massive queries or updates and are worth checking if there's apparent degraded performance on the deployment.
If you are wondering, the graph above comes from running a heavily scaled
pgbench against the database and watching the immediate aftermath.
Some of this data was in the old metric graphs, but with the new UI, the graph is updated regularly in-place. The graphs also have a wider range of viewing options from the last 30 minutes of data to all the way out to the last seven days. Dragging the mouse pointer over the graph will display the data for that moment in time making it easier to visualize the progression of memory consumption and release.
This enhancement is part of an evolution of metrics going on within Compose to ensure high quality and accessible information on all databases. Keep an eye on Compose Notes for future developments.