fix(test): do no cause secondary errors in _dump_debug_info

If firecracker already died, then the pid file might not exist anymore,
and we get a NoSuchProcess exception from thread_backtraces(). This can
be confusing when debugging failures, because its a secondary error that
only happens during error handling of the actual reason we're dumping
debug information in the first place.

Fix this by simply returning an empty list of threads in this case.

Signed-off-by: Patrick Roy <[email protected]>

Author: roypat

tests/framework/microvm.py

@@ -1101,10 +1101,12 @@ def thread_backtraces(self):
         backtraces = []
         for thread_name, thread_pids in utils.get_threads(self.firecracker_pid).items():
             for pid in thread_pids:
-                backtraces.append(
-                    f"{thread_name} ({pid=}):\n"
-                    f"{utils.check_output(f'cat /proc/{pid}/stack').stdout}"
-                )
+                try:
+                    stack = Path(f"/proc/{pid}/stack").read_text("UTF-8")
+                except FileNotFoundError:
+                    continue  # process might've gone away between get_threads() call and here
+
+                backtraces.append(f"{thread_name} ({pid=}):\n{stack}")
         return "\n".join(backtraces)
 
     def _dump_debug_information(self, what: str):

tests/framework/utils.py

@@ -34,11 +34,15 @@
 
 def get_threads(pid: int) -> dict:
     """Return dict consisting of child threads."""
-    threads_map = defaultdict(list)
-    proc = psutil.Process(pid)
-    for thread in proc.threads():
-        threads_map[psutil.Process(thread.id).name()].append(thread.id)
-    return threads_map
+    try:
+        proc = psutil.Process(pid)
+
+        threads_map = defaultdict(list)
+        for thread in proc.threads():
+            threads_map[psutil.Process(thread.id).name()].append(thread.id)
+        return threads_map
+    except psutil.NoSuchProcess:
+        return {}
 
 
 def get_cpu_affinity(pid: int) -> list: