Problem(s):

• lsof command hanging during execution of run-command-shim for 3P customers Chevron, CapGemini. This hang has caused timeouts with Run Command v1 Linux (even before executing the user's script)
  ICM incidents mentioned here Task 34627514: Investigate RCv2 Linux NOT using lsof command OR using alternatives (lsof command hanging)
• "lsof command not found" error due to its package not installed

Background:

• lsof command is being used currently in shim file of extensions (Run command v1, v2, custom script) to prevent issues caused by race conditions
• lsof command in shim is used to "list open files by the run command binary run-command-handler". If there are any open files already, we do not run the script, we wait and retry today.
• lsof is performance-intensive (lists processes, examines all file descriptors and symbolic links. It can get stuck if there are issues examining file descriptors on an unavailable/unresponsive mount, Network file system , DNS timeouts on name resolution of IPs). This can cause hang issues on high load servers of our customers

Solution

• Use flock command to execute the script (do not use lsof)
• flock command provides a way to acquire a lock on file first before executing run command script, thus preventing issues due to race conditions (note we double fork command with enable, so flock command is not blocking, it returns as soon as the background process is started to execute the script)
• Unlike lsof, which is often a standalone package, flock is part of util-linux is a core collection of essential system utilities which is almost always pre-installed on standard Linux systems. This also prevents issues such as "lsof command not found"

Sample execution from log:
2025-12-17T21:16:40.356264Z INFO ExtHandler [Microsoft.Azure.Extensions.Edp.RunCommandHandlerLinuxTest.RC1218_7-1.14.0] Command: bin/run-command-shim enable
[stdout]
flock -x -n ./run-command-handler.lock -c 'nohup /var/lib/waagent/Microsoft.Azure.Extensions.Edp.RunCommandHandlerLinuxTest-1.14.0/bin/run-command-handler enable &'
'+ flock_status=0
'+ '[' 0 -eq 1 ']'
'+ '[' 0 -eq 0 ']'
'+ LOCK_ACQUIRED=1
'+ echo 'Lock acquired on file ./run-command-handler.lock and executed command nohup /var/lib/waagent/Microsoft.Azure.Extensions.Edp.RunCommandHandlerLinuxTest-1.14.0/bin/run-command-handler enable & successfully. Exiting.'
Lock acquired on file ./run-command-handler.lock and executed command nohup /var/lib/waagent/Microsoft.Azure.Extensions.Edp.RunCommandHandlerLinuxTest-1.14.0/bin/run-command-handler enable & successfully. Exiting.
'+ break
'+ set +x