Managing MySQL in production is hard. Slow queries materialize out of nowhere. Connection pools mysteriously exhaust. Replication falls behind. Index choices that made sense six months ago are now strangling performance. And when something actually breaks at 2 AM, the evidence has usually evaporated by the time you open a terminal.

MySight was built to change that. It's a self-hosted MySQL performance monitoring platform that combines a rich, real-time dashboard with intelligent tooling: adaptive anomaly detection, automated diagnostic snapshots, query plan visualization, Slack-native alerting, and optional AI-powered analysis. Whether you're a solo DBA or a distributed engineering team, MySight gives everyone the visibility they need to move fast without breaking things.

Know Your Database's Health at a Glance

Every time you open MySight, you're greeted by a Health Score from 0 to 100 — a single composite number that tells you immediately whether your MySQL server needs attention. The score weighs four factors that matter most to production systems:

• Buffer Pool Hit Ratio (50%): Are queries being served from memory or from disk?
• P50 Response Time (25%): How fast is MySQL actually answering queries?
• Connection Utilization (15%): How close is the server to exhausting its connection limit?
• Temp Table Disk Ratio (10%): Is MySQL spilling intermediate query results to disk?

When the score drops, MySight surfaces specific recommendations right in the Info Panel — not just "something is wrong" but exactly what to fix and why.

Real-Time Monitoring That Doesn't Cost You

MySight polls your MySQL server every 10 seconds by default. Flip on Real-Time Mode during incidents or load tests and it drops to 5-second intervals, activating a live scrolling chart showing hit ratio, response time, queries/sec, and thread counts over the last 100 intervals.

For saved connections, MySight runs a shared server-side monitoring session. Ten engineers watching the same database produce the same MySQL load as one. Metrics are collected once per poll cycle and broadcast to all connected browsers simultaneously via Server-Sent Events.

If someone accidentally reloads the browser, MySight automatically restores their saved-connection sessions — no credentials stored in the browser, no re-typing connection strings.

Find and Fix Slow Queries — Fast

Query performance is usually where database problems hide. MySight gives you multiple lenses to find them:

The sys Schema Reports go even deeper: identify queries doing full table scans, queries spilling temp tables to disk, queries requiring expensive sort merge passes, and break load attribution down by MySQL user and client host. When an incident hits, you need to know which application server is responsible — MySight tells you in seconds.

InnoDB Internals, Demystified

InnoDB is a black box to most. MySight opens it up:

Panel	What You Learn
Buffer Pool	Data pages, free pages, dirty pages with a visual split layout. Hit ratio color-coded from Excellent (>99%) to Critical (<95%). Cross-references to correlated panels built in.
InnoDB I/O & Row Ops	Data reads/writes in bytes and operations. Log write volume and frequency. Row-level insert/update/delete/read rates.
Redo Log & Checkpoint	Checkpoint age as a % of redo log capacity. Current LSN, Flushed LSN, Checkpoint LSN. Warns when checkpoint lag approaches the danger zone where InnoDB stalls writes.
Tablespace & Disk	Individual tablespace files with sizes, row formats, encryption status. Per-table data/index breakdown and reclaimable space. Undo purge lag (history list length) monitoring.
Lock & Deadlock Monitor	On-demand inspection of current lock waits (blocking vs. waiting transactions) and the most recent deadlock parsed from SHOW ENGINE INNODB STATUS.

Index Intelligence

Bad indexes are expensive in two directions: missing indexes make reads slow, and redundant indexes make writes slow. MySight's Index Usage Analysis panel (on-demand, zero continuous overhead) gives you:

• Unused Indexes: Indexes that have never been used for reads since the last server start. Every unused index is costing you write performance.
• Duplicate Indexes: Indexes where one is a left-prefix of another — one of them is dead weight.
• Full Index Inventory: Every index with read/write counts and cardinality. High-write, zero-read indexes are immediate drop candidates.

The Database Tables panel complements this by flagging every table with a NO PK badge (missing primary key — causes severe InnoDB performance issues) or NO IDX badge (no indexes at all). It also shows fragmentation percentage so you know which tables are candidates for OPTIMIZE TABLE.

Connection Intelligence

Connection problems kill databases silently. MySight watches from multiple angles:

Memory Usage Breakdown

MySQL's memory usage is invisible until it causes an OOM crash. The Memory Usage panel (MySQL 8.0+) reads from performance_schema.memory_summary_global_by_event_name to show exactly how memory is allocated across InnoDB (buffer pool, AHI, log), SQL layer (parser, optimizer), temp tables, and Performance Schema.

When non-InnoDB overhead exceeds 30% of total allocation, MySight flags a potential OOM risk — typically caused by per-connection buffers (sort_buffer, join_buffer) multiplied across many connections.

Configuration Advisor: 20+ Rules, Zero Guesswork

MySight's Configuration Advisor scans SHOW GLOBAL VARIABLES and SHOW GLOBAL STATUS against 20+ expert-authored rules, flagging common MySQL misconfigurations across seven categories:

Each finding includes the current value, the specific recommendation, and the reason it matters. Severity levels guide your priorities: Critical for immediate action (data safety, major performance), Warning for near-term attention, and Info for optimization opportunities. The Configuration Advisor uses only in-memory status variables — completely safe to run on production anytime.

Historical Trends & Metric Correlation

Real-time snapshots show you now. History explains why. For saved connections, MySight stores every metric poll in PostgreSQL, building up to 30 days of time-series data.

Historical Metrics lets you plot buffer pool hit ratio, response time, query throughput, thread counts, network I/O, InnoDB row operations, and replication lag over selectable ranges: 1h, 6h, 24h, 7d, 30d. Longer ranges use server-side downsampling for performance.

Metric Correlation View lets you overlay up to 4 metrics on a shared timeline — using presets like "Response Time vs Threads Running" or "Hit Ratio vs Dirty Pages" — to discover cause-and-effect relationships. Dual Y-axes activate automatically when selected metrics have very different scales (more than 10x apart), so both remain readable.

You can also compare two saved connections side-by-side with the Server vs Server Comparison mode — primary vs replica, staging vs production. Zero additional MySQL load since it reads from stored PostgreSQL summaries.

Adaptive Anomaly Detection: Your Database Learns What Normal Looks Like

Static thresholds are blunt instruments. What's alarming at 3 AM on Sunday might be perfectly normal at noon on Monday. MySight's Adaptive Anomaly Detection computes rolling baselines from your historical data and alerts when current values deviate significantly — measured in standard deviations (σ):

MySight tracks 7 key metrics with directional awareness — it knows that a lower hit ratio is always bad, but that a drop in query volume could mean either a traffic drop or an app outage:

• Buffer Pool Hit Ratio (higher = better)
• P50 Response Time (lower = better)
• Threads Running (lower = better)
• Slow Query Count (lower = better)
• Dirty Page Ratio (lower = better)
• Total Queries (direction-neutral)
• Threads Connected (direction-neutral)

Baselines use an hour-of-day window by default, automatically upgrading to a more precise day-of-week model (e.g., "Tuesday at 2 PM") once enough samples are accumulated — adapting to your server's weekly traffic patterns.

Improvements are always capped at "info" severity regardless of deviation magnitude — so they stay visible as a confirmation that your optimization worked, without triggering false alarm fatigue.

Alert Rules + Triggered Diagnostics: Your 24/7 On-Call Teammate

MySight's Alert Rules let you define threshold conditions on 10 key metrics — buffer pool hit ratio, response time, queries/sec, threads connected, threads running, connection utilization, slow queries, temp table disk ratio, dirty pages, and replication lag. Rules evaluate client-side against already-polled data with zero additional MySQL load. When a threshold is breached:

You can also trigger a diagnostic capture manually at any time. Once stored, captures are browsable with five detail tabs — Processes, Transactions, Lock Waits, Status Diff, and InnoDB Status. Compare two captures side-by-side to see exactly what changed between the incident spike and the recovery.

This is your black box recorder. Incidents no longer have to remain mysteries because "the evidence was gone by the time I looked."

Slack-Native Sharing Across Every Panel

Every panel in MySight has a hover-revealed Share button that sends a structured Slack summary in one click. Post to a channel, or paste a Slack message link to reply directly in an existing thread. Add an optional comment before sending.

When sharing Comparison Mode to Slack, each metric is prefixed with a color indicator — 🔴 regression, 🟢 improvement, ⚪ unchanged — with values shown as baseline → current (delta%). Your teammates get full context without needing to open the dashboard.

PDF Performance Reports

Need to share findings with stakeholders or document a performance review? Export a comprehensive PDF report with 11 sections:

1. Cover page with health score visualization
2. Technical analysis — all key metrics with assessments
3. Database tables — schema sizes, fragmentation, PK coverage
4. Active processes — long-running queries, lock waits, user distribution
5. Lock & deadlock analysis
6. Query digest analytics — top patterns, no-index queries, disk temp tables
7. Configuration advisor findings by severity
8. On-demand analysis — memory, redo log, tablespace, connection pool
9. Anomaly detection & sys Schema reports
10. Non-technical summary — plain-language for management
11. Prioritized action items — ~30 categories with severity badges

Choose between an Internal Report (full technical details including server configuration and per-user attribution) and a Customer Report (sanitized version safe for external stakeholders). No third-party PDF service — generated entirely in your browser.

Built for Teams, Designed Around Security

AI-Powered Analysis: Flexible & Privacy-Respecting

Configure optional AI integration to get deeper, contextual insights on slow query logs, active processes, connection pool health, and buffer pool metrics. MySight supports 7 AI providers:

Privacy is built-in: All SQL is normalized before reaching the AI — literal values are replaced with ? so no actual data leaves your environment. Only query patterns, aggregated counters, and server metrics are sent. No credentials, no connection strings, no raw data — ever.

Monitor Multiple Servers, One Dashboard

Connect to multiple MySQL servers simultaneously using tabs. Each connection has its own independent dashboard state — metrics history, real-time mode, lock data, comparison baselines. Switching tabs preserves everything.

Inactive tabs show compact Health Summary Cards with the latest stored health score and key metrics — so you can scan all your servers at a glance without actively connecting to each one.

Works with any MySQL-compatible server: self-hosted MySQL, Amazon RDS, Amazon Aurora, Google Cloud SQL, Azure Database for MySQL, and Percona Server.

Everything MySight Monitors

Who MySight Is For