Skip to main content
Version: 0.4 · current

Ingestion modes

ConceptReviewed October 11, 2026
Applies to:Excubius 0.4Roles: Entity admin
PlansFree · Core · Pro · EnterpriseDeploymentSelf-hosted

Every source, such as New Relic, NinjaOne or Azure Monitor, reaches Excubius in one of three modes: Polling, Webhook only or Hybrid. The mode decides how fast alerts arrive, whether Excubius needs credentials for the source, and whether it gets inventory and metrics as well as alerts. This article explains each mode and when it fits.

The three modes​

ModeHow it worksStrengthsWeaknessesFits when
PollingThe collector asks the source's API on a schedule for inventory, metrics and alerts, with read-only credentialsNothing comes in from outside. It catches up after downtime. You get inventory and metricsAlerts show up within the polling interval. It uses API quotaThe source cannot reach this server: no public address, or inbound traffic is not allowed
Webhook onlyThe source pushes its alerts to a webhook URL. Nothing is polled and no credentials are storedAlerts arrive at once. No API credentials at allNo inventory and no metrics. Alerts sent while this server is down are lostYou can give the source a URL but not API credentials
HybridWebhooks for alerts, plus polling of inventory and metrics and a periodic reconciliation of alertsInstant and self-healing: an alert closed at the source while its webhook was lost is resolved by the next reconciliationSlightly more API useThe source can call this server and you can give it read-only credentials. The recommended default

You choose the mode when you add the source in Settings › Sources and can change it at any time without a restart: the collector picks up the change within 30 seconds, and webhooks are checked against the mode on every call.

What each mode does​

PollingWebhook onlyHybrid
Credentials neededYesNoYes
Inventory and metricsPolledNonePolled
AlertsPolled on the alerts intervalWebhooksWebhooks, reconciled on the reconciliation interval
Webhook callsRefused and auditedAcceptedAccepted
The source is fresh whilePolls succeedA webhook arrived in the last dayPolls succeed
TestRuns on the collectorAnswers at once from what was receivedRuns on the collector
Sync nowYesNoYes

A polled source stays fresh while it has succeeded within three of its shortest polling intervals, and at least 15 minutes. A webhook-only source can only show when it last heard something, so a quiet day turns it stale.

Reconciliation​

A reconciliation is a complete listing of the source's open alerts. An alert the source no longer lists is resolved in Excubius, with two exceptions:

  • Alerts older than the source's own listing window, 30 days for New Relic issues and Azure Monitor alerts, resolve only when the source says so.
  • Alerts someone resolved by hand in Excubius are not reopened while the source keeps reporting the same occurrence. A new occurrence opens a new alert.

Nothing is resolved from a poll that failed, and no alert is invented when a source is unreachable: the source turns stale and shows the error instead.

How webhooks are protected​

Every webhook endpoint applies the same rules, whatever the source:

  1. HTTPS and an unguessable token. The URL carries a random 256-bit token, and only a hash of it is stored. The full URL is shown once, when it is made. Rotate URL replaces it, and the old one stops working at once.
  2. Optional shared secret. When you set a Webhook secret, every call must carry it as Authorization: Bearer <secret>, as the password of Authorization: Basic, or in X-Webhook-Secret or X-Custom-Webhook-Secret. Azure Monitor action groups cannot add headers, so for them the token is the credential.
  3. Limits. The server accepts 30 webhook calls a minute per IP address, with bursts of 20, and refuses bodies over 512 KiB. An unknown token gets the same answer as a source that does not exist.
  4. Stored before processing. The payload is stored first and processed in the background. A payload that fails is kept with its error and retried three times, so nothing is lost to a bug. A retry of the same event, recognized by the source's event ID, is accepted and dropped.
  5. Audited. Calls refused because the source is in polling mode or switched off are counted on the source and audited. Every change of mode, rotation, test and sync is written to the audit log with the user and the time.

Received and ignored events are kept for 30 days; Settings › Sources shows the latest ones for each source.

Polling intervals​

Each source has its own intervals, in seconds, under Polling intervals. Leave a field empty to use its default:

SourceInventoryMetricsAlerts (polling)Alert reconciliation (hybrid)
New Relic900300300900
NinjaOne900900300900
Azure Monitor900300300900

The minimum is 60 seconds, and 300 for the reconciliation. Each check keeps its own interval too, so the metrics phase runs only the checks that are due; Key Vault expiry, for example, is read hourly. Lower intervals cost API quota: a rate limited error on the source means they are too low.

The collector's own checks do not depend on a mode: they run on the intervals of the catalog, from 30 seconds to 24 hours.