Back to Blog
2 min read

Migrating to Fabric: 5 Things I Wish I Knew

We migrated from Synapse to Fabric. It took twice as long as expected. Here’s what I wish someone had told me.

1. Start Smaller Than You Think

We tried migrating everything at once. Bad idea.

Better approach:

  • Pick one non-critical workload
  • Learn the platform
  • Build your patterns
  • Then scale

2. Your Synapse Notebooks Won’t “Just Work”

Fabric uses different runtime versions. APIs changed. Dependencies are different.

Budget time for:

  • Updating package versions
  • Fixing path references
  • Adjusting authentication
  • Testing everything

3. Direct Lake Has Requirements

Direct Lake is amazing. But it requires:

  • Data in Delta format
  • Specific data types
  • Proper partitioning
  • OneLake storage

We had to restructure data. Plan for this.

4. Monitor From Day One

Capacity metrics are crucial. Without monitoring, you’ll hit limits and not know why.

Set up alerts early:

  • Capacity utilization
  • Throttling events
  • Query failures
  • Cost tracking

5. Train Your Team

Fabric is different enough that your team needs training. Not just documentation—actual hands-on practice.

We underestimated this. Cost us 2 weeks of confusion.

The Bottom Line

Fabric is good. Migration is work. Plan accordingly.

Budget 2x your initial estimate. You’ll need it.

Michael John Peña

Michael John Peña

Senior Data Engineer based in Sydney. Writing about data, cloud, and technology.