Metrics
Lucity keeps track of how much CPU and memory each service uses, what each database is doing, and how full each volume is.
Service metrics
Open a service on the canvas and switch to the Metrics tab. It charts CPU and memory over the last hour, 6 hours, 24 hours, 7 days or 30 days.
CPU and memory over 30 days. The dashed vertical lines mark deployments.
The large number above each chart is the current usage, next to the most the service can use, which is its CPU or memory limit times the number of replicas. The dashed line along the top of the chart marks that limit, and Edit limit opens the service settings to change it. CPU and memory explains what happens when a service reaches it.
Total adds up all replicas of the service, while Per replica gives each replica its own line. Every deployment is marked with a dashed vertical line, so you can tell whether a change moved the numbers. Hovering over a chart moves a cursor across both charts, and the numbers switch to their values at that moment.
CPU is measured in cores, and anything below one core is shown in thousandths, so 250m is a quarter of a core. Metrics show up a minute or two after a service starts, and the charts refresh every 30 seconds.
Database metrics
Open a database and switch to the Metrics tab. CPU and memory sit on top, charted against the limits of the database, with a grid below them for storage, connections, transactions and rows read and written per second, the cache hit ratio, deadlocks, and how far the replica trails the primary.
CPU and memory against their limits, and what PostgreSQL is doing below them.
Every number is the current value, and the line under it puts it in context, such as the peak over the selected range or when the last deadlock happened. Hovering over any chart moves a cursor across all of them, and every number switches to its value at that moment.
Total adds up both instances, while Per instance gives each one its own line, labeled with whether it is the primary or the replica right now. A cache hit ratio close to 100% means almost every read is answered from memory, so a ratio that keeps dropping usually means the database needs more memory.
Below the charts, Top queries lists the statements that took the most time in total, with their share of that time, how often they ran, how long they took on average and how many rows they returned. Values in a statement are replaced by placeholders such as $1, so every call of the same query counts together. Open a row to read the whole statement and copy it. The list counts from the time shown next to its title and starts over when the primary changes.
Volume usage
Open a volume and switch to the Usage tab to see how much of the disk is in use, as a percentage of its size and charted over the same time ranges.
On the canvas, a mounted volume is attached under its service and fills from left to right as the disk fills up. Usage is only recorded while the volume is mounted.