Forum Discussion
Eventhouse - Engine Under Memory Pressure
- 1 year ago
Nah, sorry, the Microsoft Fabric Capacity Metric app is completely useless for troubleshooting this issue, as I mentioned in my original post. Storage and memory usage are two distinct concepts. It would be misleading to accept your reply as a solution.
I raised a support ticket with Microsoft and had a helpful engineer from Microsoft provide some useful scripts that we were able to run to gather some diagnostics on memory usage. The level of detail that's exposed to customers for KQL is apparently quite limited by design for now. I provided this feedback, which will hopefully reach the Microsoft Fabric product team for consideration. Perhaps access to more advanced monitoring & diagnostics for KQL/Evenhouse could be a product roadmap feature.
Until executing .show queries and .show commands to sift through the logs is the way.
.show commands command - Azure Data Explorer & Real-Time Intelligence | Azure Docs
Here are some of the queries that were provided by Microsoft Support team that we found helpful in identifying the root cause of memory issues:
.show queries | where StartedOn > ago(7d) | extend MemoryUsage = todouble(OverallQueryStats.resource_usage.memory.peak_per_node) | summarize totalplanning = sum(MemoryUsage) by Application, User, bin(StartedOn, 1h) | render timechart.show commands | where StartedOn > ago(7d) | extend MemoryUsage = todouble(ResourcesUtilization.MemoryPeak) | summarize totalplanning = sum(MemoryUsage) by Application, User, bin(StartedOn, 1h) | render timechart.show commands-and-queries | where StartedOn > ago(5d) | where State !="Completed".show commands-and-queries | where StartedOn > ago(7d) | where FailureReason has "Engine under memory pressure" or FailureReason has 'Low memory condition' or FailureReason has 'Query execution lacks memory resources to complete' | summarize count() by bin(StartedOn,6h) | render timechart
Hi kalinky ,
Thanks for reaching out to the Microsoft fabric community forum.
The Microsoft Fabric Capacity Metric app allows any capacity administrator to monitor OneLake Storage. Please try to monitor storage to try and check if anything specific is causing the error.
Storage is billed separately from your Fabric or Power BI Premium Capacity units. Data ingested into a KQL database is stored in two tiers of storage: OneLake Cache Storage, and OneLake Standard Storage.
If you find this post helpful, please mark it as an "Accept as Solution" and consider giving a KUDOS. Feel free to reach out if you need further assistance.
Thank you
Hi kalinky ,
May I ask if you have resolved this issue? If so, please mark the helpful reply and accept it as the solution. This will be helpful for other community members who have similar problems to solve it faster.
Thank you.