Repository navigation
Diagnostics and Performance
AE2Enhanced ships a command-driven analysis toolkit for system health checks, performance profiling, crafting-plan verification and performance anomaly alerts.
Main command:
/ae2enhanced, alias/ae2e.
Permissions
-
/ae2e perfand/ae2e planare available to everyone by default (permission level 0). Adjust withDiagnostics.analysisPermissionLevelin the config file (2= OP only); tab completion follows the same rule. -
/ae2e diag,/ae2e debugand the other maintenance subcommands require OP (permission level 2).
All diagnostic logging and data collection is switch-controlled. Changes take effect immediately at runtime, are not written back to the config file, and reset to their built-in defaults on restart.
/ae2e debug list # list all switches with their current state
/ae2e debug <switch> on|off # enable/disable (runtime override)
/ae2e debug <switch> reset # restore the built-in default
/ae2e debug <switch> status # show one switch
| Switch | Default | Purpose |
|---|---|---|
specialcrafting |
off | Verbose logs for special-recipe routing, the LP planner and gate execution |
virtualbatch |
off | Failure reasons for virtual batch crafting in the Central ME Interface |
batchaggregate |
off | Crafting CPU batch-aggregation decisions (throttled to once per 5 s) |
perf |
on | Master switch for performance collection (TPS sampling + instrumented timers). When off, /ae2e perf shows only previously collected snapshots |
planverify |
on | Crafting-plan conservation check (runs automatically on every completed plan) |
centralfluid |
off | Per-node fluid lifecycle diagnostics for the Central ME Interface (extraction → fluid-channel injection → fallbacks) |
Runs read-only health checks for the given system (default all), with results color-coded OK / WARN / ERROR in chat:
| System | What it checks |
|---|---|
storage |
Hyperdimensional storage: controller formed / safe mode (corrupted storage sections) / network active and powered; on-disk save header integrity (magic + version) |
channels |
Storage channel registration: Energy / Mana / Starlight channels against their registration conditions; gas (mekeng) and essentia (thaumicenergistics) channel presence |
recipes |
Recipe registries: Black Hole / Singularity Ritual recipe counts, Singularity Chamber index size and per-type distribution |
grid |
AE2 grids: empty grids, controller conflicts, suspected power loss (power draw with zero stored energy), crafting CPU statistics |
personaldim |
Personal dimensions: orphan entries (dimension not registered), dimension ID conflicts |
/ae2e diag check # all systems
/ae2e diag check grid # AE2 grids only
Runs all checks and writes a full report to logs/ae2enhanced/diag-<timestamp>.txt, containing:
- Environment info (mod version, relevant mod presence, switch states)
- All check results with OK / WARN / ERROR totals
- Server TPS across three windows
- Metrics registry snapshot
- Recent anomaly events (up to 50, newest first)
When reporting an issue, run this command first and attach the generated file.
/ae2e perf tps
TPS and tick-time statistics (avg / min / max / p95) over 5 s / 1 min / 2 min windows. Color-coded: green ≥ 19.5 TPS, yellow ≥ 15, red otherwise.
/ae2e perf top machine [n] # Top-N machine classes by avg tick time across all grids (default 10)
/ae2e perf top grid [n] # Top-N grids by avg tick time
/ae2e perf grid <rank> # drill into the grid at the given rank (nodes / power / storage / CPUs / controller state)
Data source: per-node tick time (nanoseconds) held in this mod's own timing registry. AE2-UEL removed the native timing field (getAvgNanoTime() always returns 0), so the mod takes over collection with a Mixin wrapping TickManager's tickingRequest calls (/ae2e debug perf off to stop it).
The average is a time-weighted EMA with a 2-second time constant, so it converges within seconds after a machine is unloaded or slowed down — a per-call EMA would barely move for machines that tick once every few seconds. Nodes that have not ticked for 30 s are treated as having no data, preventing stale values from polluting the rankings. The scan's own cost is printed at the end of each listing (overhead self-check).
/ae2e perf slow [thresholdMs] [n] # nodes whose avg tick time exceeds the threshold (default 1.0 ms, top 10)
Each line shows the machine class, average time, and coordinates + dimension so you can walk or teleport to the offender.
/ae2e perf metrics
Dumps the internal metrics registry: counters, timers (p50/p95/p99) and gauges. Examples:
- Counters —
crafting.nativeCalcAbort(native calculation aborted over budget),plan.path.<path>(executions per planner path),plan.verify.violation,plan.lp.degradedUnit - Timers —
storage.hdc.update,ae2.craftingcpu.executeCrafting,plan.computeMs.<path>,plan.computeMs.lpPhase,perf.analyzer.scan
/ae2e perf baseline set # capture the current state as the baseline (server must have been up ≥ 1 minute)
/ae2e perf baseline show # show the baseline
/ae2e perf baseline clear # delete the baseline (disables alerts)
/ae2e perf alert [multiplier] # show/set the alert multiplier (default 1.5)
Once a baseline exists, the system compares every minute: average tick time, total grid tick time, and the baseline's Top-10 machine classes. If any exceeds baseline × multiplier, an alert is broadcast to online OPs (5-minute cooldown) and recorded in the anomaly event stream (visible in diag report). The baseline persists per save (<save>/ae2enhanced/perf/baseline.json); re-run baseline set when your world enters a new stage.
/ae2e perf export
Exports TPS windows, grid/machine scans and the full metrics registry to logs/ae2enhanced/perf-<timestamp>.json for trend analysis in Excel, Grafana or any JSON-capable tool. Periodic exports form a historical trend dataset.
Effectiveness evaluation of the plan optimization strategies:
- Per-path execution counts and compute-time distribution (avg / max / p95):
special,dag,dagFallback,native - Conservation check totals: passed / violated / skipped (simulated plans)
Execution status tracking: lists all busy crafting CPUs — final output, progress percentage, co-processors and elapsed time.
Runs on every completed crafting plan (planverify switch): a non-simulated plan tree must contain zero missing materials. A violation means the planner produced an inconsistent tree; it is logged and recorded in the anomaly event stream immediately. During normal play the violation count should always be 0.
See LP Crafting Planner for how the planner itself works.
Verifies the Hyperdimensional Storage Nexus:
- Every loaded controller: structure formed, not in safe mode, network active and powered.
- Every on-disk storage entry: file header integrity — the
AE2Emagic and a valid version number. An entry whose header cannot be read is reported as an error. - Entries present on disk with no loaded controller are reported as informational, since the chunk is probably just unloaded.
Safe mode is the storage system's permanent-error state (corrupted frames, serialization failure or corruption found while loading). A section that enters safe mode becomes read-only; transient I/O faults (disk hiccups, antivirus file locks) are retried with backoff and do not trigger it.
Verifies personal dimension bookkeeping:
- An entry whose dimension ID is assigned but not registered → ERROR (orphan entry).
- An entry whose dimension ID is registered but is not a personal dimension type → ERROR (ID conflict).
- An unassigned, empty entry (rules edited ahead of time) is normal and does not warn.
A storage stress test: inject count randomly generated enchanted gear types (swords, tools, armor, bows with 0–5 random enchantments and random large stack sizes) into the storage of the formed Hyperdimensional Controller with the given UUID. count accepts 1 to 100,000.
This is how the storage system is exercised against the "tens of thousands of distinct item types" scenario. Use /ae2e recoverhd list to find a UUID, then run the test against it.
The following events are recorded automatically (256-entry ring buffer, newest first):
- Native crafting calculation aborted over budget (default 30 s budget)
- Special-recipe detector analysis over budget (conservative miss)
- Plan conservation violations
- Performance anomaly alerts
View them in the "recent anomaly events" section of a /ae2e diag report.
| Collection | Mechanism | Cost |
|---|---|---|
| TPS | One nanoTime diff per tick |
Negligible |
| Grid / machine times | Mixin-wrapped tickingRequest timing with a time-weighted EMA (AE2-UEL removed native timing) |
Two nanoTime calls per node tick, same as original AE2 behavior |
| L2/L3 instrumentation (HDC update, executeCrafting) | 1-in-20 sampling + nanoTime
|
Negligible |
| Plan verification | One tree walk per completed plan | Proportional to crafting request rate, very low |
| Alert comparison | One full grid scan per minute | Same as a single /ae2e perf top
|
Stop all collection at once: /ae2e debug perf off.