Fixed resource monitoring

src/ptbench/utils/resources.py

@@ -342,8 +342,10 @@ def _monitor_worker(
             ra.acc()  # guarantees at least an entry will be available
 
             if summary_event.is_set():
-                queue.put(ra.summary())
+                summary = ra.summary().copy()
+                queue.put(summary)
                 ra.clear()
+                print(queue.get())
                 summary_event.clear()
 
             time.sleep(interval)
@@ -416,7 +418,6 @@ class ResourceMonitor:
         """Forces the monitoring process to yield data and clear the internal
         accumlator."""
         self.summary_event.set()
-
         try:
             data: dict[str, list[int | float]] = self.q.get(
                 timeout=2 * self.interval

[André Anjos]

@dcarron: not sure what happened here, but this stopped working for me. Why are you copying the summary? This is an interprocess communication queue - the data is always copied.

[André Anjos]

Ah - was the summary container being deleted by the Q?

[Daniel CARRON]

Yes, the self.q.get() in checkpoint() was getting a summary with all metrics set to 0, although the values were set when it was put inside the queue.

[André Anjos]

You also seem to have left a print() statement dangling here.

[André Anjos]

Removing the print statement makes the thing work for me again.

[Daniel CARRON]

Sorry about the print statement, sometimes vscode is acting weird when trying to save something when ssh-ing from home.

[André Anjos]

It is alright.