Loki logs sending instance="none" by default

[candlerb]

Required information

• Distribution: Ubuntu
• Distribution version: 22.04.4
• Incus version: 6.0-202404040310-ubuntu22.04

Issue description

The documentation for the config setting loki.instance says:

Key: loki.instance
Type: string
Default: Local server host name or cluster member name
Scope: global

This allows replacing the default instance value (server host name) by a more relevant value like a cluster identifier.

On my servers, it is unset:

root@nuc3:~# incus config get loki.instance

root@nuc3:~# hostname
nuc3
root@nuc3:~#

However, the logs are being sent with instance="none" rather than instance="<hostname>"

root@loki:~# logcli query '{app="incus"}' --since=1h --limit=3
2024/04/17 09:14:27 http://localhost:3100/loki/api/v1/query_range?direction=BACKWARD&end=1713345267346769206&limit=3&query=%7Bapp%3D%22incus%22%7D&start=1713341667346769206
2024/04/17 09:14:27 Common labels: {app="incus", instance="none", location="none", type="logging"}
                                                 ^^^^^^^^^^^^^^^
2024-04-17T08:41:41Z {} level="info" Done updating images
2024-04-17T08:41:41Z {} level="info" Done pruning expired backups
2024-04-17T08:41:41Z {} level="info" Updating images

WORKAROUND: set loki.instance explicitly on each host.

incus config set loki.instance=nuc3

Steps to reproduce

1. Install Loki
2. Forward logs using loki.api.url. Don't set loki.instance
3. Examine logs and look at values of instance label