Your data lives in your Poets, on your own disks. Each kind of data is kept for a set time, and a Poet also has a disk limit it never passes. You change both in Settings › Retention and disk; each setting shows, right beside it, how much that kind of data takes now, how far back it reaches and how it has grown over the last days.

What is kept, and for how long

A week of logs and traces matches how long an incident keeps its full detail, so last week reads the same whether or not an incident was open. Expired data cannot be brought back, not even by rolling a setting back.

The disk limit

A Poet uses at most its disk limit: what you set in Most disk per Poet (or POET_DISK_LIMIT when you install it), else its own data plus the free space left on its volume. When it passes 90 % of the limit, it frees room until it is back at 80 %, oldest first, in this order:

1. metrics at full detail that are already kept at 10 s;
2. log lines and traces;
3. other metrics at full detail;
4. metrics at 10 s, then at 1 min, then at 10 min;
5. deploys and other changes.

Dashboards, incidents and your settings are never removed to make room. Both percentages are settings too. A volume shared with other programs that fill it makes the Poet give back a share of its oldest data, never all of it.

Seeing it

Each Poet reports what it keeps, as part of its own dashboard (“This Poet” › Storage): the disk and memory each kind of data takes, how far back each reaches, the disk limit, and what was removed and why (its age, or the disk limit).