Metrics, logs and limits#
Recent invocations#
Functions → a function → Invocations and logs, or
The last 200 calls the cluster relayed: the invocation's id, the target and
version, the trigger (api, http, schedule, tool), the instance, when,
how long the caller waited (a cold start and a wait in the queue included),
the outcome (ok, error, timeout, rejected, unavailable) and the
status. Never the input or the answer. They are kept in memory: a restart of
the cluster forgets them. Calls made at your edge machines' URLs are not in
this list; they are in the metrics below.
Logs of one invocation#
Every line a call writes (print, console.log, a traceback) is prefixed
with its invocation's id, [inv-…]. The id is in the answer's
x-astraeus-invocation-id header.
Logs stay on the machines: they are read from the instances' machines when
you ask, and an instance that has stopped took its log with it. Without
--invocation, the instances' latest lines.
Metrics#
Functions → a function → Metrics, or astra astraeus functions metrics
weather, or GET /v1/functions/{name}/metrics:
| Number | What |
|---|---|
invocations, errors, timeouts |
Calls each instance served since it started, and how many failed or ran past the timeout. |
rejected |
Calls an instance refused because all its slots were taken. |
in_flight / capacity |
Calls running now, and how many the instances serve at once. |
p50_ms, p95_ms, p99_ms |
Each instance's durations over its last 1024 calls; the totals give the slowest instance's. |
| Waiting, refused (busy) | Per target, as the cluster relayed them: calls in the queue now, and calls refused for a full queue. |
The instances' numbers are scraped from them every 15 s, as any run's metrics; an instance's counts start again when it does.
Limits#
| What | Limit |
|---|---|
| Name | lower-case letters, digits and -, starting with a letter; 30 characters |
| Code | one file, 256 KiB |
| Packages | 100, each a line of at most 200 characters, no options |
| Variables / Credentials | 64 / 16 |
| Timeout | 1–900 s (default 30) |
| Memory | 128–65536 MiB (default 512) |
| Input | 6 MiB (API, CLI, edge URLs); 1 MB at the hosted URL |
| Calls at once per instance | 1–1000 (default 8) |
| Instances | 1–64 (default 4) |
| Queue | 0–10000 calls (default 64) |
| Aliases | 32 |
| Cold start | a call waits at most 120 s for an instance |
| Hosted URL | 90 s per call |
| Recent invocations | the last 200, in memory |