Profiler HTML output exceeds browser rendering capacity on complex pages

Question

ContextMagento 2's built-in profiler (Magento\Framework\Profiler) can be enabled via bin/magento dev:profiler:enable html_file to generate flame-graph HTML reports in var/profiler.*.html. The profiler captures timer IDs, elapsed time, memory usage, and call counts for each instrumented code block.ProblemOn complex pages (e.g., category listings with layered navigation, checkout with many quote items), the generated HTML file frequently exceeds 50 MB. Browsers struggle to parse and render such large single-page flame graphs, resulting in frozen tabs, excessive memory consumption, or outright crashes before analysis can begin. The profiler does not implement pagination, streaming, or incremental rendering for the HTML driver.ConstraintsCSV output remains parseable at any size but lacks the visual flame-graph view.Custom drivers can stream to external systems (Blackfire, Tideways), yet the built-in drivers provide no such integration.Magento recommends against enabling the profiler in production due to overhead and potential exposure of internal timing data, but no enforced guard exists.What configuration or workflow allows analyzing bottlenecks on heavy pages without hitting the browser rendering limit? Is there a supported method to split or stream the HTML profiler output? Can the CSV data be reliably transformed into a navigable flame graph using offline tooling?

Tasadduq Burney · Answer

Recommended workflow

Enable the CSV profiler driver, convert the CSV to a flame‑graph offline, and only fall back to HTML splitting if a visual flame‑graph is required.

Steps

Enable CSV profiler: bin/magento dev:profiler:enable csv_file
Generate the page request (e.g., via curl or browser) to produce var/profiler.*.csv.
Verify the CSV size is manageable (should stay under a few MB even for heavy pages).
If the CSV includes parent_timer (or a similar hierarchy column), convert it to a collapsed stack format:

# Example: produce "timer;elapsed_time" lines
awk -F, '{print $2";"$4}' var/profiler.*.csv > collapsed.txt
# Adjust column numbers if your CSV differs

Feed the collapsed file to a flame‑graph generator, e.g., Brendan Gregg’s flamegraph.pl:

perl flamegraph.pl collapsed.txt > profiler.svg

Or import the CSV into SpeedScope (File → Import → CSV) and select the timer and elapsed‑time columns.

When HTML visualisation is still needed

If you must keep the HTML driver, split the generated file after collection:

Rename the HTML file to profiler_full.html.
Use a simple script to wrap each top‑level <tr> block into its own HTML page with shared CSS/JS:

awk '/^<tr class="top"/ {if (out) close(out); out=sprintf("part%02d.html",++n); print "<!DOCTYPE html><html><head><link rel=\"stylesheet\" href=\"profiler.css\"></head><body>" > out; print $0 > out; next}
     {print $0 > out}' profiler_full.html

Adjust the pattern to match your actual markup; duplicate any required CSS/JS files or reference them via relative paths.

Missing diagnostic detail

Confirm whether your CSV export contains a parent_timer column (or equivalent hierarchy). If it does not, you will need to reconstruct the call stack from timer IDs and timestamps before converting to a flame‑graph.

Profiler HTML output exceeds browser rendering capacity on complex pages

Context

Problem

Constraints

1 answer

Recommended workflow

Steps

When HTML visualisation is still needed

Missing diagnostic detail

Your perspective belongs here.

0 question comments