commit 522cd98aecaf678c29206cf56db1f1fb17fd27d1
author Congbin Guo <guocb@google.com> Thu Oct 06 11:47:28 2022 -0700
committer Chromeos LUCI <chromeos-scoped@luci-project-accounts.iam.gserviceaccount.com> Tue Oct 11 23:15:37 2022 +0000
tree 57059a17a0b8f0936aeedadbfa334d25e2442cb5
parent 9f7517aef3542212c0ad9474ebf91129a836460f

chore(sysmon): update process names to monitor

We add more chrome fleet services to monitor, e.g. fleet-tlw,
common-tls, etc. Also, we deprecated a legacy service, i.e.
gs_archive_server from the list.

BUG=b:238798008
TEST=unit test

Change-Id: I5157193b8fce42887f3288ed64ff037d41a1f31c
Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/chromite/+/3937607
Commit-Queue: Congbin Guo <guocb@chromium.org>
Auto-Submit: Congbin Guo <guocb@chromium.org>
Tested-by: Congbin Guo <guocb@chromium.org>
Reviewed-by: Allen Li <ayatane@chromium.org>

scripts/sysmon/proc_metrics.py

diff --git a/scripts/sysmon/proc_metrics.py b/scripts/sysmon/proc_metrics.py
index 0b537b6..b54f765 100644
--- a/scripts/sysmon/proc_metrics.py
+++ b/scripts/sysmon/proc_metrics.py
@@ -35,13 +35,22 @@
     def __init__(self):
         self._metrics = [
             _ProcessMetric("autoserv", test_func=_is_parent_autoserv),
+            _ProcessMetric(
+                "common-tls", test_func=partial(_is_process_name, "common-tls")
+            ),
             _ProcessMetric("curl", test_func=partial(_is_process_name, "curl")),
             _ProcessMetric(
-                "getty", test_func=partial(_is_process_name, "getty")
+                "dnsmasq", test_func=partial(_is_process_name, "dnsmasq")
             ),
             _ProcessMetric(
-                "gs_archive_server",
-                test_func=partial(_is_python_module, "gs_archive_server"),
+                "drone-agent",
+                test_func=partial(_is_process_name, "drone-agent")
+            ),
+            _ProcessMetric(
+                "fleet-tlw", test_func=partial(_is_process_name, "fleet-tlw")
+            ),
+            _ProcessMetric(
+                "getty", test_func=partial(_is_process_name, "getty")
             ),
             _ProcessMetric(
                 "gs_offloader",
@@ -50,6 +59,9 @@
             _ProcessMetric("gsutil", test_func=_is_gsutil),
             _ProcessMetric("java", test_func=partial(_is_process_name, "java")),
             _ProcessMetric(
+                "labservice", test_func=partial(_is_process_name, "labservice")
+            ),
+            _ProcessMetric(
                 "lxc-attach", test_func=partial(_is_process_name, "lxc-attach")
             ),
             _ProcessMetric(
@@ -61,6 +73,7 @@
                 "sysmon",
                 test_func=partial(_is_python_module, "chromite.scripts.sysmon"),
             ),
+            _ProcessMetric("tko_proxy", test_func=_is_tko_proxy),
         ]
         self._other_metric = _ProcessMetric("other")
 
@@ -171,3 +184,19 @@
         and cmdline[0] == "python"
         and cmdline[1].endswith("gsutil")
     )
+
+
+def _is_tko_proxy(proc):
+    """Return whether proc is a tko proxy.
+
+    A tk proxy process is like
+    '/opt/cloud_sql_proxy -dir=<...>
+        -instances=google.com:chromeos-lab:us-central1:tko
+        -credential_file=<...>'.
+    """
+    cmdline = proc.cmdline()
+    return (
+        len(cmdline) == 4
+        and cmdline[0].split("/")[-1] == 'cloud_sql_proxy'
+        and cmdline[2] == '-instances=google.com:chromeos-lab:us-central1:tko'
+    )