InfluxDB Operations

Commands below use the 2.x CLI. Confirm the equivalents for your major version before running anything — see the version note in InfluxDB Data Model.

Ingestion

influx write --bucket metrics --precision s --file batch.lp

Batching is essential: thousands of points per request rather than one point per request. Telegraf is the usual collection agent and buffers on the client side, which protects against a backend restart.

Practical guidance:

• Use the coarsest timestamp precision your data justifies.
• Sort points by series where the client makes that easy; it improves compression.
• Handle write rejections explicitly — a silently dropped batch is data loss.

Retention and downsampling

influx bucket create --name metrics_raw   --retention 7d
influx bucket create --name metrics_daily --retention 730d

A scheduled task aggregates raw data into the longer-retention bucket:

option task = { name: "downsample-hourly", every: 1h }
 
from(bucket: "metrics_raw")
  |> range(start: -2h)
  |> filter(fn: (r) => r._measurement == "readings")
  |> aggregateWindow(every: 1h, fn: mean)
  |> to(bucket: "metrics_daily")

Clustering

Clustering has differed by version and edition. The open-source distribution has historically been single-node, with horizontal scaling available in commercial offerings.

Backup and restore

influx backup /backups/influx-$(date -u +%Y%m%dT%H%M%SZ) --token $INFLUX_TOKEN
 
influx restore /backups/influx-20260731T020000Z --token $INFLUX_TOKEN
# Restore one bucket into a new name to verify without touching production.
influx restore /backups/influx-20260731T020000Z \
  --bucket metrics --new-bucket metrics_verify --token $INFLUX_TOKEN

Monitoring

InfluxDB exposes Prometheus-format metrics at /metrics.

Watch:

• Series cardinality per bucket — the leading indicator of trouble. See Cardinality.
• Write and query request rates, errors and durations.
• Task execution success and last-run time for every downsampling task.
• Disk usage for the data directory and the WAL, with a growth projection.
• Memory usage, which tracks the index and therefore cardinality.
• Dropped or rejected writes, which are data you did not store.

influx bucket list
influx task list
influx task retry-failed --id <task-id>

Alert on cardinality growth rate rather than only on an absolute threshold — a doubling in a day is a deploy that added a tag, and catching it early is far cheaper than recovering afterwards.