Observability

Know which step was slow. Not just that the run was.

Logs, metrics, and traces in one place, emitted as OpenTelemetry from the free tier up. When an integration takes four seconds instead of one, the trace already shows you which call spent the time.

  • Metrics
  • Traces
  • Alerts
  • Selective logging
See it on your workflowsWhen something fails
Trace
payroll-sync · production
  • payroll-sync4.21s
  • Schedule12ms
  • Query · employees460ms
  • Data Mapper240ms
  • REST Client · Workday2.6s
  • Upsert · NetSuite510ms
  • Send To Engage230ms
62% of this run was spent in one call to Workday.
OpenTelemetry

Three signals, one standard, every plan

Logs, metrics, and traces are emitted as OpenTelemetry — the open standard your monitoring stack already speaks. Point them at the tools you have, or use the dashboards built in. Available from the free tier, not reserved for an enterprise contract.

Metrics

Is it healthy?

Throughput, latency, and error rates per integration and per tenant. The numbers that tell you something changed before anyone files a ticket.

Traces

Where did the time go?

Each run broken into spans, with the duration of every activity. A slow integration stops being a mystery and becomes a specific call to a specific system.

Logs

What actually happened?

Structured output from every step, at the level you chose, routed to the logger you named — not one undifferentiated stream.

Because it is OpenTelemetry rather than a proprietary format, none of this locks you in. The same data flows to whatever you already use to watch the rest of your estate.

Incident management

Nobody watches a dashboard at 3am

Dashboards answer questions you thought to ask. Alerts tell you about the ones you did not. Koodisi pushes the failure to the people who own it, through the channels and the service desk they already work in.

Alerts
Be told when a run fails or a deployment does not land, rather than finding out when someone downstream notices the data is wrong.
Escalation
A failure nobody acknowledges should not sit unowned. Escalation moves it along until somebody picks it up.
ITSM integration
Route incidents into the service desk your organisation already runs, so integration failures follow the same process as everything else.

Incident management is available on paid plans.

Alert routing
payroll-sync · production
When

A run fails in production

  • On-callIntegration platform team
    Immediately
  • Service deskIncident raised in ITSM
    Immediately
  • EscalationEngineering manager
    Unacknowledged 15m
Escalation continues until somebody acknowledges the failure.
Selective logging

Everything logged is nothing found

A log that records every step at every level is as unhelpful during an incident as no log at all. You choose what each step writes, where it goes, and which values are reduced on the way.

LevelsDebugInfoWarnError

Log at the level the environment needs

Run Debug while you are building and drop to Warn in production, without editing the workflow's logic to do it.

Route to named loggers

Give a Log step a logger name and its output separates from everything else, so noisy diagnostics from one integration stay out of the stream your on-call team watches.

Mask the field, keep the record

Fields you designate as sensitive are reduced before they are written — an account number down to its last four digits — while the rest of the payload stays readable enough to debug.

FAQ

Questions about observability

What is emitted, who gets told, and where it stops.

Still have a question? Talk to an expert.

Seeing the failure is half of it.

Koodisi Engage owns the other half — retry, fallout management, and somewhere for a person to resolve what could not clear itself.

How recovery works