Memory leak if asitop kept running for a few minutes

[jvicenti]

If asitop is kept running for a while it will eventually consume all of the memory on the system, showing in Activity Monitor as Python. Even after a few minutes the memory usage steadily increases from 30MB to a few GB according to the "Real Memory"

[SlinkierElm5611]

I am noticing this behaviour too.

[ap4499]

Noticed this too, I had a look at the source code, and I think this should work:
sudo asitop --max_count 600

This forces PowerMetrics to restart after X polls.

600 can of course be any integer value. That would force a restart at around 10 minutes given the poll time is 1s. Moving it lower might be desirable (although there is a small trade off in the PowerMetrics reboot time/cost if taken to an extreme).

The default value is 0, which will in effect cause the Powermetrics data to grow infinitely, which will have the following effects:

• Large temp files.
• Unbounded/increasing memory usage
• Increased CPU usage

[prosorin]

Fix Available: Memory Leak Root Cause + Solution

I've identified and fixed the root cause of this memory leak (also affects #81, #71, #83).

🔍 Root Cause

The issue is in asitop/utils.py:14 - the code reads the entire temp file into memory on every iteration:

with open(path+timecode, 'rb') as fp:
    data = fp.read()  # ← Reads ENTIRE file into memory

What happens:

1. powermetrics continuously appends to /tmp/asitop_powermetricsXXXXXX
2. File grows indefinitely (I observed 3.6GB after 39 hours)
3. asitop reads entire file every second but only uses the last plist entry
4. Memory usage grows proportionally to file size

Result: 20GB memory usage after 39 hours in my testing

✅ Solution

Read only the last 2MB of the temp file instead of the entire file:

def parse_powermetrics(path='/tmp/asitop_powermetrics', timecode="0"):
    data = None
    try:
        with open(path+timecode, 'rb') as fp:
            # Memory optimization: read only the last 2MB instead of entire file
            # This prevents memory bloat when temp file grows to GB sizes over long runs
            fp.seek(0, 2)  # Seek to end
            file_size = fp.tell()
            read_size = min(file_size, 2 * 1024 * 1024)  # Read max 2MB
            fp.seek(max(0, file_size - read_size))  # Seek to starting position
            data = fp.read()
        data = data.split(b'\x00')
        powermetrics_parse = plistlib.loads(data[-1])
        # ... rest unchanged

📊 Results

Runtime	Before Fix	After Fix
1 hour	~200MB	~50MB
39 hours	~20GB	~50MB
1 week	~80GB+	~50MB

✅ Testing

• macOS Sonoma 14.x
• Apple Silicon (M1/M2/M3)
• asitop 0.0.24
• Tested over 39+ hours continuous operation

Before fix: 20GB memory, 3.6GB temp file
After fix: 50MB memory (same 3.6GB temp file, but only reads last 2MB)

🎯 Impact

This 5-line fix solves:

• ✅ Memory leak if asitop kept running for a few minutes #80 - Memory leak
• ✅ Big disk space consumption in /private/tmp/asitop_powermetricsXXX #81 - Disk space consumption impact
• ✅ /tmp file grows quite large and is left after quit? #71 - Temp file growth impact
• ✅ After two days running the app crashes #83 - Crashes after extended runtime

Fully backward compatible, no API changes.

Happy to submit a PR if this approach looks good to the maintainer!

asitop_memory_fix.patch

--- a/asitop/utils.py
+++ b/asitop/utils.py
@@ -10,8 +10,13 @@ import plistlib
 def parse_powermetrics(path='/tmp/asitop_powermetrics', timecode="0"):
     data = None
     try:
         with open(path+timecode, 'rb') as fp:
-            data = fp.read()
+            # Memory optimization: read only the last 2MB instead of entire file
+            # This prevents memory bloat when temp file grows to GB sizes over long runs
+            fp.seek(0, 2)  # Seek to end
+            file_size = fp.tell()
+            read_size = min(file_size, 2 * 1024 * 1024)  # Read max 2MB
+            fp.seek(max(0, file_size - read_size))  # Seek to starting position
+            data = fp.read()
         data = data.split(b'\x00')
         powermetrics_parse = plistlib.loads(data[-1])
         thermal_pressure = parse_thermal_pressure(powermetrics_parse)

[zachtheyek]

it doesn't seem like the repo is actively maintained anymore. @prosorin should we consider a fork of asitop to implement your changes? the current iteration is insanely resource hungry and it seems to me like an obvious design flaw to read the entire temp file into memory without any periodic garbage collection, or any periodic cleanup of the existing temp file to reduce disk pressure.

[jvicenti]

I agree with @zachtheyek that this doesn't appear to be maintained any more so it should be forked. Also I don't see a reason why there needs to be a temp file that grows indefinitely as well.

[prosorin]

ok.. I'll do what I can and let you know! thanks for the feedback.

[zachtheyek]

@prosorin i don't mind forking the repo myself if you don't want to be an active maintainer beyond your initial change?

edit: I'll fork and takeover ownership later this month. Give me some time y'all. Will announce here + create a separate issue so people know to migrate if they want up-to-date patches