Skip to main content

Limits & Quotas

xScaler enforces these limits per tenant.

Limits table​

LimitValue
Max active series per tenantPlan-dependent
Max labels per series30
Max label name length1,024 characters
Max label value length2,048 characters
Max metric name length1,024 characters
Max samples per remote_write request2,000
Query timeout2 minutes

:::info Plan-dependent limits Limits marked "Plan-dependent" vary by your subscription tier. Log in to the xScaler dashboard and navigate to Settings → Limits to see the exact values for your account. :::

Max samples per request​

The ingest endpoint accepts a maximum of 2,000 samples per request. This is reflected in the queue_config.max_samples_per_send setting in both Prometheus and Grafana Alloy configurations. The default value in those configs is already set to 2000. Do not raise it above this limit.

Query timeout​

All queries time out after 2 minutes. If your query hits this limit:

  1. Narrow the time range.
  2. Increase the step value to reduce the number of data points evaluated.
  3. Add label selectors to cut the number of series scanned.

Finding high-cardinality metrics​

Labels with unbounded values (user IDs, request IDs, session tokens) drive unexpected series growth. This query finds the top offenders:

curl "https://euw1-01.m.xscalerlabs.com/api/v1/query" \
  -H "Authorization: Bearer <token>" \
  -H "X-Scope-OrgID: <tenant-id>" \
  --data-urlencode 'query=topk(10, count by (__name__)({__name__=~".+"}))'

When one metric dominates the count, check its labels and drop the ones carrying unbounded values.

Reducing ingest volume​

To cut series before you hit the limit:

  • Drop unused metrics at the source. Use write_relabel_configs in Prometheus or Alloy to drop metrics before they leave the scrape host (see Prometheus remote_write).
  • Aggregate before shipping. Use write_relabel_configs or Alloy prometheus.relabel blocks to pre-aggregate high-cardinality metrics and ship only the aggregated series.
  • Review label cardinality. Replace high-cardinality label values (e.g. pod IDs) with coarser ones (e.g. deployment name).