[BUG] High Memory using Full Dashboard

[CADelete]

Component

Full Dashboard

Performance Monitor Version

High Memory Usage and Not Releasing

SQL Server Version

SQL Server 2019 CU32 GDR

Windows Version

Windows Server 2019 Datacenter

Describe the Bug

I am seeing high memory usage when using the Full Dashboard.

It has gone as high as 4GB and doesn't seem to release without the process being killed. Even when the dashboard is closed but still running in the background Memory continues to grow. I am monitoring 1 staging environment which is a busy server.

I also notice when this happens the refresh button isn't available in the dashboard, but you can move between tabs and change the duration

Steps to Reproduce

Open Dashboard (Run as Administrator)
Connect to a server
Run processes to simulate high CPU, locking etc
Watch Memory increase and not release

Expected Behavior

I'd expect Memory to not grow that high and also release when not in use

Actual Behavior

Memory fluctuates but stays high, randomly increasing. Refresh Tab button is not available

Error Messages / Log Output

Log Output below

Screenshots

Additional Context

SQL Server agent is running, Developer Edition in use. It is an Azure VM with SQL Server 2019 installed

[CADelete]

When killing the process, opening the dashboard again and then immediately closing (keeping the process running, just not having the dashboard open), memory stays low, around 50MB. As soon as I open the GUI it again and open the server tab, memory increases back to 4GB.

Schedule is set to balanced, changing to low impact has no effect

[erikdarlingdata]

Fix shipped in PR #836 (merged to dev)

Root cause: The Dashboard had the same memory leak patterns that were fixed in Lite (commit 2b61592) but never received those fixes. When server tabs were opened, 53+ event handlers and chart hover helpers were subscribed but never unsubscribed on tab close, preventing GC from collecting the tab and its child controls.

What was fixed

• Added Dispose() to Dashboard's ChartHoverHelper — unsubscribes MouseMove/MouseLeave handlers
• Added DisposeChartHelpers() to ServerTab, MemoryContent, ResourceMetricsContent, and QueryPerformanceContent
• Stored lambda event delegates as named fields so they can be unsubscribed in ServerTab_Unloaded
• Unsubscribed all child control events on tab close (slicers, CriticalIssuesTab, MemoryTab, PerformanceTab, ResourceMetricsContent)
• Stored and unsubscribed AlertAcknowledged handler in MainWindow.CloseTab_Click
• Cleared unfiltered data collections, filter/legend dictionaries, and baseline provider cache on tab close

60-minute monitoring results (dev build, post-fix)

Test config: 2 server tabs open (RDS + SQL2022), RDS tab in focus with active auto-refresh and chart interaction.

Time	Dashboard Memory
09:14	363 MB (start)
09:23	420 MB (peak)
09:35–09:55	346–363 MB (stable)
10:15	192 MB (Gen 2 GC reclaimed)
10:20–10:25	188 MB (stable post-GC)

Start: 363 MB → End: 188 MB (48% reduction over 60 min). No upward memory trend despite continuous server tab activity. GC successfully reclaims memory that was previously pinned by leaked event handlers.

This will be included in the next release.

[CADelete]

Installed v2.7.0. Issue still occurring.

Initially when opening the Overview tab is is stable at 50MB. Opening a server, initially Memory hits around 180MB but after a minute CPU and Memory Increases. Refresh Tab is greyed out. Closing the tab makes no difference, Memory continues to rise until I have to kill the process.

No issues with the FinOps tab, memory is stable when using that.

[CADelete]

From the Log

[erikdarlingdata]

@CADelete Thanks for following up — sorry this is still happening. The v2.7.0 fix (#836) addressed event handler leaks but clearly there's another cause.

A couple questions to help us pin this down:

1. Which sub-tab are you on when the memory starts climbing? (Overview, Queries, Resource Metrics, Memory, Locking, System Events, or Plan Viewer?)
2. Auto-refresh interval — what's it set to in Settings? (e.g., 30s, 60s, etc.)
3. How many servers do you have tabs open for when this happens?
4. The log screenshot is a bit small — could you paste the text of the error messages, or share a higher-res version?
5. Does it still happen if you disable auto-refresh (toggle it off in the toolbar) before opening a server tab?

That last question will tell us whether this is the auto-refresh loop or the initial load itself.

[CADelete]

I'm on the Overview tab
Auto refresh is set to off. In settings it is set to 30 seconds
I only have one server installed

Log File Info

[2026-04-16 10:45:14.099] [ERROR] Error refreshing QueryPerformance data: Execution Timeout Expired. The timeout period elapsed prior to completion of the operation or the server is not responding.
Exception: SqlException
Message: Execution Timeout Expired. The timeout period elapsed prior to completion of the operation or the server is not responding.
StackTrace: at Microsoft.Data.SqlClient.SqlConnection.OnError(SqlException exception, Boolean breakConnection, Action1 wrapCloseInAction) at Microsoft.Data.SqlClient.SqlInternalConnection.OnError(SqlException exception, Boolean breakConnection, Action1 wrapCloseInAction)
at Microsoft.Data.SqlClient.TdsParser.ThrowExceptionAndWarning(TdsParserStateObject stateObj, SqlCommand command, Boolean callerHasConnectionLock, Boolean asyncClose)
at Microsoft.Data.SqlClient.SqlCommand.EndExecuteReaderInternal(IAsyncResult asyncResult)
at Microsoft.Data.SqlClient.SqlCommand.EndExecuteReaderAsync(IAsyncResult asyncResult)
at System.Threading.Tasks.TaskFactory1.FromAsyncCoreLogic(IAsyncResult iar, Func2 endFunction, Action1 endAction, Task1 promise, Boolean requiresSynchronization)
--- End of stack trace from previous location ---
at PerformanceMonitorDashboard.Services.DatabaseService.GetQueryStatsAsync(Int32 hoursBack, Nullable1 fromDate, Nullable1 toDate, Boolean fromSlicer) in D:\a\PerformanceMonitor\PerformanceMonitor\Dashboard\Services\DatabaseService.QueryPerformance.cs:line 1180
at PerformanceMonitorDashboard.Services.DatabaseService.GetQueryStatsAsync(Int32 hoursBack, Nullable1 fromDate, Nullable1 toDate, Boolean fromSlicer) in D:\a\PerformanceMonitor\PerformanceMonitor\Dashboard\Services\DatabaseService.QueryPerformance.cs:line 1227
at PerformanceMonitorDashboard.Controls.QueryPerformanceContent.RefreshAllDataAsync(Boolean fullRefresh) in D:\a\PerformanceMonitor\PerformanceMonitor\Dashboard\Controls\QueryPerformanceContent.xaml.cs:line 970
[2026-04-16 10:45:16.683] [INFO] === Application Exiting (Exit Code: 0) ===

[2026-04-16 11:43:11.157] [ERROR] Error refreshing QueryPerformance data: Execution Timeout Expired. The timeout period elapsed prior to completion of the operation or the server is not responding.
Exception: SqlException
Message: Execution Timeout Expired. The timeout period elapsed prior to completion of the operation or the server is not responding.
StackTrace: at Microsoft.Data.SqlClient.SqlConnection.OnError(SqlException exception, Boolean breakConnection, Action1 wrapCloseInAction) at Microsoft.Data.SqlClient.SqlInternalConnection.OnError(SqlException exception, Boolean breakConnection, Action1 wrapCloseInAction)
at Microsoft.Data.SqlClient.TdsParser.ThrowExceptionAndWarning(TdsParserStateObject stateObj, SqlCommand command, Boolean callerHasConnectionLock, Boolean asyncClose)
at Microsoft.Data.SqlClient.SqlCommand.EndExecuteReaderInternal(IAsyncResult asyncResult)
at Microsoft.Data.SqlClient.SqlCommand.EndExecuteReaderAsync(IAsyncResult asyncResult)
at System.Threading.Tasks.TaskFactory1.FromAsyncCoreLogic(IAsyncResult iar, Func2 endFunction, Action1 endAction, Task1 promise, Boolean requiresSynchronization)
--- End of stack trace from previous location ---
at PerformanceMonitorDashboard.Services.DatabaseService.GetQueryStatsAsync(Int32 hoursBack, Nullable1 fromDate, Nullable1 toDate, Boolean fromSlicer) in D:\a\PerformanceMonitor\PerformanceMonitor\Dashboard\Services\DatabaseService.QueryPerformance.cs:line 1180
at PerformanceMonitorDashboard.Services.DatabaseService.GetQueryStatsAsync(Int32 hoursBack, Nullable1 fromDate, Nullable1 toDate, Boolean fromSlicer) in D:\a\PerformanceMonitor\PerformanceMonitor\Dashboard\Services\DatabaseService.QueryPerformance.cs:line 1227
at PerformanceMonitorDashboard.Controls.QueryPerformanceContent.RefreshAllDataAsync(Boolean fullRefresh) in D:\a\PerformanceMonitor\PerformanceMonitor\Dashboard\Controls\QueryPerformanceContent.xaml.cs:line 970

The auto refresh is set to off

If I open the server and then close it immediately, the same thing happens, so it doesn't appear to be linked to the tab actually being open, the Memory grows regardless as long as it is has been opened.

[erikdarlingdata]

Thanks for the details — the log errors are very helpful. Here's what we've found:

Root cause: When you open a server tab, the Dashboard fires all 6 tab refreshes in parallel (Overview, Queries, Resource Metrics, Memory, Locking, System Events) — even though you're only looking at Overview. The Queries tab refresh runs GetQueryStatsAsync, which is the heaviest query we have — it groups all collected query stats by hash and includes full execution plan XML. On your server it's timing out after 120 seconds, meaning it's running for 2 full minutes pulling data before giving up.

During those 2 minutes, SqlDataReader is streaming rows into memory — each one potentially carrying hundreds of KB of plan XML. Multiply that across multiple parallel queries all hitting the database at once, and you get the memory climb you're seeing. Since there's no cancellation support, closing the tab doesn't stop the in-flight queries either.

Planned fix: Stop loading all tabs in parallel on open. Only load the visible tab, and lazy-load others when you switch to them. This is the real fix — your Overview tab loads fine at 50MB, so you should never be waiting 2 minutes for Queries data you haven't asked to see yet.

One question for us: Could you tell us a bit about your monitoring database? Specifically:

• How many servers are you monitoring with the Dashboard?
• Roughly how large is the PerformanceMonitor database? (SELECT name, size * 8 / 1024 AS size_mb FROM sys.database_files)
• How long has data been collecting? (days/weeks/months)
• Is the query_stats table particularly large? (SELECT COUNT(*) FROM dbo.query_stats)

We're curious whether this is a "lots of data" situation or a "query needs optimization" situation (or both). Either way, the lazy-load fix will prevent the problem entirely.

[CADelete]

I am only monitoring 1 server, it has 49 databases on it. The monitor is on a stand alone server which has 8GB of Memory

PerformanceMonitor 5120MB
PerformanceMonitor_log 320MB

I have been collecting data for 3 days

Query stats table has 742,228 records