KAFKA-18343: Use java_pids to implement pids (#18312)

Related to KAFKA-18343, Currently, there is an issue that the ps ax output is truncated, which causes the Kafka process ID to be unavailable. This issue can be mitigated by replacing ps ax with jcmd (i.e. using java_pids in ducktape), as it does not suffer from the truncation problem. Reviewers: Justine Olshan <[email protected]>
apache · Dec 24, 2024 · 6635d82 · 6635d82
1 parent fb7a6fb
commit 6635d82
Showing 1 changed file with 2 additions and 4 deletions.
diff --git a/tests/kafkatest/services/kafka/kafka.py b/tests/kafkatest/services/kafka/kafka.py
@@ -951,9 +951,7 @@ def run_features_command(self, op, feature, new_version):
     def pids(self, node):
         """Return process ids associated with running processes on the given node."""
         try:
-            cmd = "ps ax | grep -i %s | grep -v grep | awk '{print $1}'" % self.java_class_name()
-            pid_arr = [pid for pid in node.account.ssh_capture(cmd, allow_fail=True, callback=int)]
-            return pid_arr
+            return node.account.java_pids(self.java_class_name())
         except (RemoteCommandError, ValueError) as e:
             return []
 
@@ -1935,4 +1933,4 @@ def get_offset_shell(self, time=None, topic=None, partitions=None, topic_partiti
         return output
 
     def java_class_name(self):
-        return "kafka.Kafka"
+        return "kafka\.Kafka"