Has Anyone Faced Intermittent Query Failures When Using Reflection-Accelerated Datasets in Dremio Cloud?

Hello

I am currently testing a setup on Dremio Cloud where we rely heavily on raw & aggregation reflections to speed up BI dashboards (especially with Power BI and Tableau). :slightly_smiling_face: While the performance gains are great; I have started noticing a strange issue; some queries that normally use reflections intermittently fail with a vague “Reflection not available” error; even though the reflection status shows as accelerated and healthy.:innocent:

A simple retry often resolves it; but this is obviously not ideal for automated dashboards.:thinking:

I have verified that the datasets haven’t changed, the reflections have not been rebuilt recently & the jobs page doesn’t show any reflection refresh conflicts. :thinking: My suspicion is that something in Dremio Cloud’s internal caching or reflection routing logic might be causing temporary unavailability. :innocent:

Has anyone else experienced this? I am curious if there’s a best practice around staggering reflection refreshes / using fallback strategies to avoid this inconsistency.:thinking:

If this sounds familiar or you’ve solved something similar, I would love to hear your insights. In the meantime; I found this Dremio documentation on Reflection Refresh Best Practices that helped clarify how refresh timing works but it doesn’t cover transient failures well. :thinking: Since Dremio Cloud operates entirely in the cloud; it’s a great case study when explaining what Is Cloud Computing in real-world analytics workflows.

Any additional recommendations for stable use in production dashboards would be appreciated.

Thank you !!:slightly_smiling_face:

@ricaxa Even if reflection is not available, query should not fail. Was the reflection created just then or has been in CAN_ACCELERATE status on sys.reflections?

On the job profile under the acceleration page, what does it say?