Fabric's workspace monitoring runs up a bill of its own. It writes logs into a Monitoring Eventhouse, and that Eventhouse consumes the same capacity units your pipelines run on, at Microsoft's published pay-as-you-go rate of roughly $0.18 per CU-hour, plus OneLake storage for the logs. Azure Data Factory's Monitor view is free, but it keeps 45 days of history and answers "did the pipeline run?" rather than "which of the forty steps inside it is stuck?". Faced with paying for the better view in capacity units, some organizations simply turn it off. We've watched them do it.
How does Fabric charge for monitoring?
When you enable workspace monitoring, Fabric creates a Monitoring Eventhouse in your workspace, with an Eventstream feeding it. Both are ordinary Fabric items. They consume capacity units like any other workload, and they keep metering as long as logs flow. You can see the charge in the Capacity Metrics app under the item names "Monitoring Eventhouse" and "Monitoring Eventstream". Log storage bills separately through OneLake, at about $0.246 per GB-month cached and $0.023 per GB-month standard under current published rates.
The structure matters more than any single rate. The monitor and the workloads it watches draw from one pool. You sized your F-SKU for pipelines, notebooks, and reports, and monitoring adds a permanent draw against that same pool.
What does that look like in dollars?
Capacity pricing is public. Pay-as-you-go runs about $0.18 per CU-hour in US regions, which puts an always-on F2 near $263 a month and an F64 near $8,410. Reservations bring both down by roughly 40%. Every capacity unit that monitoring consumes is bought at your SKU's rate, and the bigger exposure sits at the SKU boundaries. F-SKUs double (2, 4, 8, up through 64 and 128), so when a monitoring draw helps push a nearly full capacity over the line, what you actually pay for is the jump to the next SKU.
Monitoring options for an ADF / Fabric estate. Directional. Verify rates for your region and SKU.
Runs inside the ADF service. Keeps 45 days of history, answers at the run level, and needs Azure portal access most of the team doesn't have.
Runs as an Eventhouse on your capacity. Bills at your SKU's rate, meters continuously, and competes with the workloads it watches.
Runs in your Azure subscription. You pay for ingestion and retention, and you build and maintain the queries, dashboards, and alerts yourself.
Runs outside the capacity, over orchestration metadata. Zero CUs, history for as long as your database keeps it, and readable without portal access.
Isn't ADF's monitor free? Why isn't that enough?
It's free, and for a small estate it's genuinely useful. Three limits show up as an estate grows. Data Factory keeps pipeline-run data for 45 days, and a query further back returns nothing. The Monitor view answers at the pipeline level, so when one parent pipeline wraps forty inner activities, "In progress" says nothing about whether the numbers will be fresh by 8 a.m. And reading it requires Azure portal access, which most of the people asking "is the data in yet?" don't have.
Keeping history past 45 days means routing diagnostic logs to Azure Monitor or Log Analytics. That works. It also adds per-gigabyte ingestion charges, plus the engineering time to build the queries and dashboards that ADF didn't ship with.
Why do teams turn Fabric's monitoring off?
Because it competes with their workloads for capacity. In our engagements we've watched organizations disable workspace monitoring after sizing and budgeting the capacity for workloads alone. Once monitoring consumed enough units to notice, they had to choose between watching the pipelines and running them. Running won. On a shared pool that's the rational choice, even though it leaves the estate blind.
What's the alternative for an ADF or Fabric estate?
Put the watching somewhere the running can't crowd it out. When orchestration is metadata, with processes, schedules, phases, and current state stored as rows in a database you own, then "what's running, what failed, what resumed" is a report over those rows. That's how ETLMaestro's status report works. It reads orchestration metadata rather than your capacity, so it consumes zero CUs. History lasts as long as your database keeps rows, and anyone who can open a report can read it, with no portal access and no 45-day limit. Pricing is published, with a free two-month trial, and the wider argument for metadata-driven orchestration is in the master pipeline problem.
Where this applies: the comparison above is about orchestration-level visibility for teams consolidating data on ADF or Fabric. If you need deep engine telemetry (query plans, capacity throttling analysis, per-visual Power BI performance), Fabric's native tooling does things a metadata report doesn't try to do. Plenty of estates run both, with the expensive one scoped to the engineers who read it.
Sources and verification
Figures verified 2026-08-12 against the sources below. This page is reviewed quarterly. Rates vary by region and change over time, so check the pricing pages before budgeting.
- Microsoft Learn: Workspace monitoring overview (billing as capacity consumption; Monitoring Eventhouse item names; storage rates)
- Microsoft Fabric blog: Announcing activation of billing for workspace monitoring
- Azure pricing: Microsoft Fabric capacity (pay-as-you-go and reservation rates by SKU)
- Microsoft Learn: Monitor Data Factory programmatically (45-day pipeline-run retention)
- Microsoft Learn: Monitor Azure Data Factory (routing diagnostics to Azure Monitor / Log Analytics for longer retention)