This is a lightweight app that cleans old data from prometheus pushgateway. You simple define the expiration duration and the pushgateway-cleaner takes care of the rest. Pushgateway-cleaner looks for metric groups that were last updated more that CLEANER_EXPIRATION_DURATION ago and deletes them.
If you require this tool 99.9% prometheus-pushgateway is not used as expected, but unfort in practice there are those 0.1% of the cases that the cost/effort of fixing this just doesn't worth it. This 0.1% is the reason this tool exists.
docker pull konnoska/pushgateway-cleaner:latest
Pushgateway-cleaner is configured entirely through environmetn variables.
| Variable | Default | Description |
|---|---|---|
| CLEANER_EXPIRATION_DURATION | 24h | Metric groups whose last push timestamp is older that this value are deleted. Accepted format {hours}h{Minutes}m{Seconds}s i.e. 1h30m20s, 1h, 10m |
| CLEANER_CLEANING_INTERVAL | 12h | Defines how often to start the cleaning routine. |
| CLEANER_ENDPOINT | localhost:9091 | Prometheus' pushgateway endpoint. |
| CLEANER_LOG_LVL | INFO | One of DEBUG, INFO, ERROR, WARNING. |
| CLEANER_DRY_RUN | "FALSE" | Set to "TRUE" to dry-run and see which metric groups would be deleted. |
If you run prometheus-pushgateway in k8s, we propose using the cleaner as a sidecar. The community's helm chart has builtin support for sidecars, so adding the following configuration to pushgateway's chart should be enough.
extraContainers:
- name: pushgateway-cleaner
image: konnoska/pushgateway-cleaner:v1.0.0
env:
- name: CLEANER_EXPIRATION_DURATION
value: "10m"