Merged
Conversation
gh release create fires both created and published events, causing two identical builds that conflict on asset uploads. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…igger Fix duplicate release builds: trigger on published only
PlanAnalyzer (Dashboard + Lite): - Rules 3 & 20: SubTreeCost threshold 0.01 → 1.0 (CTFP is integer) - Rule 3: Smarter MaxDOPSetToOne — 3-branch logic with query text check - Rule 5: AllocatesResources gate on zero-rows warning (Hash/Sort/Spool) - Rule 11: Enriched scan-with-predicate (cost %, elapsed %, selectivity) - Rule 12: IsFunctionOnColumnSide — fix non-SARGable false positive - Rules 25/31: GetWaitStatsAdvice for targeted parallelism advice - New Rule 32: Scan cardinality misestimate (actual plans) - New Rule 33: CE guess detection (estimated plans) ShowPlanParser (Dashboard + Lite): - Fix multi-statement batch parsing (iterate all <Statements> children) - Add synthetic root nodes for DECLARE/ASSIGN statements Closes #816. Syncs from PerformanceStudio PR #213. Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
FilterUpgrades used `FromVersion >= current` which excluded upgrades when the user was on a patch release (e.g., 2.4.1 skipped the 2.4.0-to-2.5.0 upgrade because 2.4.0 < 2.4.1). Changed to `ToVersion > current` so any upgrade the user hasn't reached is included. Adds regression test for the patch version scenario. Fixes #817 Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
When no new deadlocks occurred for a database, the MERGE source was empty and no row was created, leaving stale delta values in the last row. Expanded the MERGE source with a combined_source CTE that carries forward databases from the previous collection with zero counts, so the delta calculation properly resets to 0. Fixes #803 Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
#821) Adds an opt-in checkbox in both Dashboard and Lite Add/Edit Server dialogs that sets MultiSubnetFailover=true on the connection string. Recommended for AG listeners and FCIs spanning multiple subnets — connects to all IPs in parallel during failover instead of sequentially. Defaults to off; persists to servers.json; backward compatible with existing configs. Fixes #813 Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…822) The installer now checks for pre-downloaded SQL files in a community/ directory before downloading from GitHub. Users in air-gapped environments can place the files there manually. If a file is missing, it falls back to the normal GitHub download. Expected files: - community/sp_WhoIsActive.sql - community/DarlingData.sql - community/Install-All-Scripts.sql Fixes #814 Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…823) Queries sys.dm_os_host_info (SQL 2017+) for the host_distribution value. Falls back to parsing @@Version for SQL 2016 and Azure SQL DB. No schema changes — the data is queried live alongside existing server properties. Fixes #748 Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…s to Logical CPUs The timer tick handler set _isRefreshing = true before calling RefreshAllDataAsync, which also checked _isRefreshing and returned immediately — every auto-refresh was a no-op. Removed the redundant guard from the tick handler so RefreshAllDataAsync owns the flag. Renamed "CPUs" to "Logical CPUs" in the FinOps server inventory tab (both Dashboard and Lite) to clarify the relationship with the Cores/Socket column, which shows physical cores. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…s to Logical CPUs (#825) The timer tick handler set _isRefreshing = true before calling RefreshAllDataAsync, which also checked _isRefreshing and returned immediately — every auto-refresh was a no-op. Removed the redundant guard from the tick handler so RefreshAllDataAsync owns the flag. Renamed "CPUs" to "Logical CPUs" in the FinOps server inventory tab (both Dashboard and Lite) to clarify the relationship with the Cores/Socket column, which shows physical cores. Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…indow Replace the connection-success MessageBox popup with inline status in the DatabaseStatusPanel so the entire flow (connection result, database detection, install prompt) reads naturally in one place. Add guidance text explaining what Install Now does for first-time users. Size the dialog to the screen work area height and top-align it so the post-test content is never hidden below the fold. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…g-ux Improve Add Server dialog UX
…itor (#828) 01_widen_version_columns.sql (PR #722) was missing USE PerformanceMonitor and SET options that every other upgrade script has. The installer connects to master, so the script's OBJECT_ID check resolved against master, found nothing, skipped both ALTERs, and reported success while leaving columns at nvarchar(255). Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Adds _template.sql with the required SET options and USE PerformanceMonitor preamble so future upgrade scripts don't repeat the #828 mistake. Updates the README example to match the actual pattern used by all upgrade scripts. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…ng-use-828 Fix upgrade script executing against master instead of PerformanceMonitor
The three heaviest MCP tools (get_top_queries_by_cpu, get_top_procedures_by_cpu, get_query_store_top) timed out on large databases because they scanned every row, decompressed text on every row, and returned all results for client-side filtering. Dashboard: add new MCP-specific service methods using multi-phase temp tables — aggregate numerics first, rank TOP N, then hydrate text via OUTER APPLY for only the winners. DECOMPRESS now runs on ~N rows instead of the entire table. Lite: rewrite queries with CTE-based phasing — inner CTE aggregates numeric-only columns with LIMIT, outer query hydrates text via LEFT JOIN LATERAL for winners. Both: push database_name filter and TOP N into SQL instead of client-side filtering. Shared UI methods are untouched. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…zation-826 Optimize MCP query tools for large databases
…ataAsync (#833) Same class of bug as Lite (#824) — the timer tick handlers each managed the _isRefreshing flag redundantly. On a transient connection failure, MessageBox.Show blocked the UI thread while _isRefreshing remained true, causing every subsequent tick to skip silently. - Move _isRefreshing ownership into LoadDataAsync (matches Lite pattern) - Remove redundant guard from all 3 timer tick handlers - Suppress MessageBox during auto-refresh (fullRefresh: false) — log errors and update status bar instead of blocking the UI thread Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…#834) The _isRefreshing guard in LoadDataAsync could get permanently stuck if any SQL query hung (connection drop, slow timeout), silently killing all subsequent auto-refresh ticks. Replaced with a stop/start pattern: the timer stops before each tick and restarts in finally, bypassing LoadDataAsync entirely for auto-refresh. This guarantees the timer always restarts regardless of what happens during the refresh. Also removed the redundant _isRefreshing guard from CorrelatedTimelineLanesControl.RefreshAsync which was independently causing Server Trends charts to silently skip updates. Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
… on tab close (#835) Dashboard had the same memory leak patterns fixed in Lite (commit 2b61592) but never received those fixes. When server tabs were closed, 53+ event handlers and hover helpers remained subscribed, preventing GC from collecting the tab and its child controls. - Add Dispose() to Dashboard ChartHoverHelper (matches Lite) - Add DisposeChartHelpers() to ServerTab, MemoryContent, ResourceMetricsContent, and QueryPerformanceContent - Store lambda event delegates as named fields so they can be unsubscribed - Unsubscribe all child control events in ServerTab_Unloaded (slicers, CriticalIssuesTab, MemoryTab, PerformanceTab, ResourceMetricsContent) - Store and unsubscribe AlertAcknowledged handler in MainWindow.CloseTab_Click - Clear unfiltered data collections, filter dictionaries, legend panels, and baseline provider cache on tab close Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…-leaks Fix Dashboard memory leaks on tab close (#835)
- Bump version to 2.7.0 in all 4 csproj files - Add CHANGELOG entry for v2.7.0 (features, changes, fixes) - Update README: MultiSubnetFailover AG note, air-gapped community scripts - Replace DispatcherTimer auto-refresh with async Task.Delay loop to prevent priority starvation under heavy UI load (chart rendering) Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Release v2.7.0
…gend duplication Replace DispatcherTimer with async Task.Delay loop to prevent priority starvation under heavy UI load. Don't cancel the loop on Unloaded (WPF fires Unloaded on tab switch/close, not just control destruction). Catch OperationCanceledException from SQL queries without killing the loop. Skip auto-refresh ticks while a full refresh is in progress to prevent concurrent chart rendering that duplicates legends. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…h-resilience Fix Dashboard auto-refresh: async loop, tab close, legend duplication
…leanupOnClose WPF fires Unloaded on tab switch, not just destruction. The old handler tore down legend tracking, chart helpers, and event subscriptions on every tab switch, causing the auto-refresh loop to lose track of legend panels and create duplicates. Now Unloaded is a no-op; full cleanup only runs via CleanupOnClose when a tab is actually closed/removed. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…leanup Fix legend duplication on tab switch
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Added
community/directory for air-gapped environments ([FEATURE] - Have cached copies of tools included #814, Add offline community script support via community/ directory (#814) #822)Changed
Fixed
Test plan
🤖 Generated with Claude Code