Do "lnet stop" instead of lustre_rmmod (#80)

* Do "lnet stop" instead of lustre_rmmod This should fix the hanging references from LU-11986 Signed-off-by: Nathaniel Clark <[email protected]> * Fixup lnet start/stop Signed-off-by: Nathaniel Clark <[email protected]> * Update chroma_agent/action_plugins/manage_lnet.py * Adjust stop/unload to not relay on sbin/lnet /usr/sbin/lnet is not present on clients. lustre_rmmod ptlrpc is not affected by LU-11986 Signed-off-by: Nathaniel Clark <[email protected]> * fixups Signed-off-by: Joe Grund <[email protected]> * fixup logic error Signed-off-by: Joe Grund <[email protected]> * fix another logic bug Signed-off-by: Joe Grund <[email protected]> * load existing modprobe files Signed-off-by: Joe Grund <[email protected]>
AlexTalker · Apr 24, 2019 · d1deaa3 · d1deaa3
1 parent 07e24fc
commit d1deaa3
Showing 1 changed file with 4 additions and 12 deletions.
diff --git a/chroma_agent/action_plugins/manage_lnet.py b/chroma_agent/action_plugins/manage_lnet.py
@@ -35,11 +35,8 @@ def start_lnet():
     """
     console_log.info("Starting LNet")
 
-    # modprobe lustre is a hack for HYD-1263 - Fix or work around LU-1279 - failure trying to mount
-    # should be removed when LU-1279 is fixed
-    return agent_ok_or_error(
-        AgentShell.run_canned_error_message(["lctl", "net", "up"])
-        or AgentShell.run_canned_error_message(["modprobe", "lustre"])
+    return AgentShell.run_canned_error_message(
+        ["lnetctl", "lnet", "configure", "--all"]
     )
 
 
@@ -50,15 +47,10 @@ def stop_lnet():
     """
 
     console_log.info("Stopping LNet")
-    # FIXME: Due to LU-11986 (partial lustre_rmmod leads to panic) and LU-9525 ("lctl
-    # network down" won't work...) call lustre_rmmod to shutdown network
 
-    # Teardown to ksocklnd - ignore error
-    result = AgentShell.run(["lustre_rmmod"])
-    if result.rc == 0:
-        return agent_result_ok
     return agent_ok_or_error(
-        AgentShell.run_canned_error_message(["lctl", "network", "down"])
+        AgentShell.run_canned_error_message(["lustre_rmmod", "ptlrpc"])
+        or AgentShell.run_canned_error_message(["lnetctl", "lnet", "unconfigure"])
     )