Merge pull request 'refs #2592 fix syntax' (#53) from fix into main

Reviewed-on: #53

CHANGELOG.md

@@ -6,6 +6,41 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [7.3.2] - 2025-08-04
+
+### Fixed
+
+- Fix syntax
+
+## [7.3.1] - 2025-08-04
+
+### Fixed
+
+- On user logout, write the user to the log
+- On EjecutarScript with client=true, return a job ID
+
+## [7.3.0] - 2025-07-31
+
+### Fixed
+
+- Wait for zombies
+
+### Changed
+
+- Change "jobid" for "job_id" for consistency
+
+## [7.2.2] - 2025-07-30
+
+### Added
+
+- Add missing file stop-agent.ps1
+
+## [7.2.1] - 2025-07-29
+
+### Changed
+
+- Run process in a new POSIX session and group, send termination signals to the whole process group
+
 ## [7.2.0] - 2025-07-28
 
 ### Added

linux/debian/changelog

@@ -1,3 +1,35 @@
+ogagent (7.3.2-1) stable; urgency=medium
+
+  * Fix syntax
+
+ -- OpenGnsys developers <info@opengnsys.es>  Mon, 04 Aug 2025 14:36:47 +0200
+
+ogagent (7.3.1-1) stable; urgency=medium
+
+  * On user logout, write the user to the log
+  * On EjecutarScript with client=true, return a job ID
+
+ -- OpenGnsys developers <info@opengnsys.es>  Mon, 04 Aug 2025 14:22:52 +0200
+
+ogagent (7.3.0-1) stable; urgency=medium
+
+  * Wait for zombies
+  * Change "jobid" for "job_id" for consistency
+
+ -- OpenGnsys developers <info@opengnsys.es>  Thu, 31 Jul 2025 10:35:16 +0200
+
+ogagent (7.2.2-1) stable; urgency=medium
+
+  * Add missing file stop-agent.ps1
+
+ -- OpenGnsys developers <info@opengnsys.es>  Wed, 30 Jul 2025 10:20:21 +0200
+
+ogagent (7.2.1-1) stable; urgency=medium
+
+  * Run process in a new POSIX session and group, send termination signals to the whole process group
+
+ -- OpenGnsys developers <info@opengnsys.es>  Tue, 29 Jul 2025 12:52:08 +0200
+
 ogagent (7.2.0-1) stable; urgency=medium
 
   * Log whether we are in ogLive or not

src/VERSION

@@ -1 +1 @@
-7.2.0
+7.3.2

src/opengnsys/ipc.py

@@ -110,6 +110,11 @@ class ClientProcessor(threading.Thread):
         logger.debug('Got Client message {}={}'.format(msg, REV_DICT.get(msg)))
         if self.parent.clientMessageProcessor is not None:
             self.parent.clientMessageProcessor(msg, data)
+        if msg == REQ_LOGIN:
+            if b',' in data:
+                self.user = data.split (b',')[0]
+            else:
+                self.user = data
 
     def run(self):
         self.running = True
@@ -197,8 +202,14 @@ class ClientProcessor(threading.Thread):
                 logger.error('Invalid message in queue: {}'.format(e))
 
         logger.debug('Client processor stopped')
-        if os.path.exists ('/windows/temp'): open ('/windows/temp/ogagentuser_died', 'w').close()
-        else:                                open (         '/tmp/ogagentuser_died', 'w').close()
+        if os.path.exists ('/windows/temp'):
+            fd = open ('/windows/temp/ogagentuser_died', 'wb')
+            fd.write (self.user)
+            fd.close()
+        else:
+            fd = open ('/tmp/ogagentuser_died', 'wb')
+            fd.write (self.user)
+            fd.close()
         try:
             self.clientSocket.close()
         except Exception:

src/opengnsys/jobmgr.py

@@ -23,32 +23,32 @@ class JobMgr():
         logger.debug ('args "{}"'.format (args))
         now = datetime.now (tz=timezone.utc)
         ts = now.strftime ('%Y-%m-%d %H:%M:%S.%f%z')    ## '%s' doesn't work on windows
-        jobid = hashlib.sha256 (now.isoformat().encode('UTF-8') + script.encode ('UTF-8')).hexdigest()[0:12]
+        job_id = hashlib.sha256 (now.isoformat().encode('UTF-8') + script.encode ('UTF-8')).hexdigest()[0:12]
         p = subprocess.Popen (args, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
-        self.jobs[jobid] = { 'p': p, 'pid': p.pid, 'starttime': ts, 'script': script, 'client': is_client, 'status': 'running', 'stdout': '', 'stderr': '' }
-        self.jobs[jobid]['t1'] = threading.Thread (target=job_readstdout, args=(self.jobs[jobid],))
-        self.jobs[jobid]['t2'] = threading.Thread (target=job_readstderr, args=(self.jobs[jobid],))
-        self.jobs[jobid]['t1'].start()
-        self.jobs[jobid]['t2'].start()
+        self.jobs[job_id] = { 'p': p, 'pid': p.pid, 'starttime': ts, 'script': script, 'client': is_client, 'status': 'running', 'stdout': '', 'stderr': '' }
+        self.jobs[job_id]['t1'] = threading.Thread (target=job_readstdout, args=(self.jobs[job_id],))
+        self.jobs[job_id]['t2'] = threading.Thread (target=job_readstderr, args=(self.jobs[job_id],))
+        self.jobs[job_id]['t1'].start()
+        self.jobs[job_id]['t2'].start()
         logger.debug ('jobs "{}"'.format (self.jobs))
-        return jobid
+        return job_id
 
     def prepare_jobs(self):
         ## can't return self.jobs because the Popen object at self.jobs[id]['p'] is not serializable. So, need to create a new dict to return
         st = []
-        for jobid in self.jobs:
-            j = self.jobs[jobid]
+        for job_id in self.jobs:
+            j = self.jobs[job_id]
             entry = dict ((k, j[k]) for k in ['pid', 'starttime', 'script', 'client', 'status', 'stdout', 'stderr'])
-            entry['jobid'] = jobid
+            entry['job_id'] = job_id
             if j['p'].poll() is not None:    ## process finished
                 entry['rc'] = j['p'].returncode
                 entry['status'] = 'finished'
             st.append (entry)
         return st
 
-    def terminate_job(self, jobid):
-        if jobid not in self.jobs: return {}
-        p = self.jobs[jobid]['p']
+    def terminate_job(self, job_id):
+        if job_id not in self.jobs: return {}
+        p = self.jobs[job_id]['p']
         p.terminate()
         time.sleep (1)
         if p.poll() is not None:

src/opengnsys/linux/OGAgentService.py

@@ -74,10 +74,12 @@ class OGAgentSvc(Daemon, CommonService):
             while self.isAlive:
                 client_died=False
                 if os.path.exists ('/tmp/ogagentuser_died'):
+                    with open ('/tmp/ogagentuser_died', 'rb') as fd:
+                        u = fd.read()
                     os.unlink ('/tmp/ogagentuser_died')
                     client_died=True
                 if client_died:
-                    self.notifyLogout (b'')
+                    self.notifyLogout (u)
 
                 # In milliseconds, will break
                 self.doWait(1000)

src/opengnsys/modules/client/OpenGnSys/__init__.py

@@ -50,13 +50,13 @@ class OpenGnSysWorker(ClientWorker):
     def process_script(self, json_params):
         script = json_params['code']
         logger.debug('Processing message: script({})'.format(script))
-        self.jobmgr.launch_job (script, True)
-        #self.sendServerMessage('script', {'op', 'launched'})
+        job_id = self.jobmgr.launch_job (script, True)
+        self.sendServerMessage('script_launched', {'op': 'launched', 'job_id': job_id})
 
     def process_terminatescript(self, json_params):
-        jobid = json_params['jobid']
-        logger.debug('Processing terminatescript request, jobid "{}"'.format (jobid))
-        self.jobmgr.terminate_job (jobid)
+        job_id = json_params['job_id']
+        logger.debug('Processing terminatescript request, job_id "{}"'.format (job_id))
+        self.jobmgr.terminate_job (job_id)
 
     def process_preparescripts(self, json_params):
         logger.debug('Processing preparescripts request')

src/opengnsys/modules/server/OpenGnSys/__init__.py

@@ -369,24 +369,46 @@ class OpenGnSysWorker(ServerWorker):
         logger.debug('received script "{}"'.format(script))
 
         if post_params.get('client', 'false') == 'false':
-            jobid = self.jobmgr.launch_job (script, False)
-            return {'op': 'launched', 'jobid': jobid}
+            job_id = self.jobmgr.launch_job (script, False)
+            return {'op': 'launched', 'job_id': job_id}
 
         else:   ## post_params.get('client') is not 'false'
             ## send script as-is
             self.sendClientMessage('script', {'code': script})
-            #return {'op': 'launched', 'jobid': jobid}       ## TODO obtain jobid generated at the client (can it be done?)
-            return {'op': 'launched'}
+
+            ## wait for job_id generated at the client
+            job_id = None
+            iters = 0
+            while True:
+                time.sleep (0.2)
+                if os.path.exists ('/tmp/EjecutarScript-jobid'):
+                    with open ('/tmp/EjecutarScript-jobid', 'r') as fd:
+                        job_id = fd.read()
+                    break
+                iters += 1
+                if iters >= 10: break
+
+            try: os.unlink ('/tmp/EjecutarScript-jobid')
+            except: pass
+
+            if job_id is None: return {'op': 'launched'}
+            else:              return {'op': 'launched', 'job_id': job_id}
+
+    def process_client_script_launched(self, data):
+        fd = open ('/tmp/EjecutarScript-jobid', 'w')
+        fd.write (data['job_id'])
+        fd.close()
+        return True
 
     @execution_level('full')
     @check_secret
     def process_terminatescript(self, path, get_params, post_params, server):
-        jobid = post_params.get('jobid', None)
-        logger.debug('Processing terminate_script request, jobid "{}"'.format (jobid))
-        if jobid is None:
+        job_id = post_params.get('job_id', None)
+        logger.debug('Processing terminate_script request, job_id "{}"'.format (job_id))
+        if job_id is None:
             return {}
-        self.sendClientMessage('terminatescript', {'jobid': jobid})
-        self.jobmgr.terminate_job (jobid)
+        self.sendClientMessage('terminatescript', {'job_id': job_id})
+        self.jobmgr.terminate_job (job_id)
         return {}
 
     @execution_level('full')

src/opengnsys/windows/OGAgentService.py

@@ -108,10 +108,12 @@ class OGAgentSvc(win32serviceutil.ServiceFramework, CommonService):
             while self.isAlive:
                 client_died=False
                 if os.path.exists ('/windows/temp/ogagentuser_died'):
+                    with open ('/windows/temp/ogagentuser_died', 'rb') as fd:
+                        u = fd.read()
                     os.unlink ('/windows/temp/ogagentuser_died')
                     client_died=True
                 if client_died:
-                    self.notifyLogout (b'')
+                    self.notifyLogout (u)
 
                 # Pumps & processes any waiting messages
                 pythoncom.PumpWaitingMessages()

src/opengnsys/workers/oglive_worker.py

@@ -35,6 +35,7 @@ import re
 import time
 try: import dbus     ## don't fail on windows (the worker will later refuse to load anyway)
 except: pass
+import select
 import random
 import subprocess
 import threading
@@ -158,10 +159,10 @@ class ogLiveWorker(ServerWorker):
             if 'thread' not in self.thread_list[job_id]: return { 'res': 2, 'der': 'Job is not running' }
             t   = self.thread_list[job_id]['thread']
             pid = self.thread_list[job_id]['child_pid']
-            logger.debug (f'pid ({pid})')
+            logger.debug (f'pid/pgid/sid ({pid})')
             try_times = 8
             sig = signal.SIGTERM
-            msg = f'could not kill pid ({pid}) after ({try_times}) tries'
+            msg = f'could not killpg pid ({pid}) after ({try_times}) tries'
             success = 2    ## mimic cmd['res'] in respuestaEjecucionComando(): "1" means success, "2" means failed
             while True:
                 t.join (0.05)
@@ -175,10 +176,10 @@ class ogLiveWorker(ServerWorker):
                 ## this is fine in the first iteration of the loop, before we send any signals. In the rest of iterations, after some signals were sent, msg should be 'job terminated' instead.
                 if pid:
                     if os.path.exists (f'/proc/{pid}'):
-                        logger.debug (f'sending signal ({sig}) to pid ({pid})')
+                        logger.debug (f'sending signal ({sig}) to process group ({pid})')
                         ## if the process finishes just here, nothing happens: the signal is sent to the void
-                        os.kill (pid, sig)
-                        #subprocess.run (['kill', '--signal', str(sig), str(pid)])
+                        os.killpg (pid, sig)
+                        #subprocess.run (['kill', '--signal', str(sig), f'-{pid}'])   ## negative PID is used for sending signals to the process group
                     else:
                         msg = f'pid ({pid}) is gone, nothing to kill'
                         success = 1
@@ -283,31 +284,35 @@ class ogLiveWorker(ServerWorker):
                 proc = ['bash', '-c', '{} {}'.format (devel_bash_prefix, exe)]
 
         logger.debug ('subprocess.run ("{}")'.format (' '.join (proc)))
-        p = subprocess.Popen (proc, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
+        p = subprocess.Popen (proc, stdout=subprocess.PIPE, stderr=subprocess.PIPE, start_new_session=True)
         if self.pid_q:
-            self.pid_q.put (p.pid)
+            self.pid_q.put (p.pid)     ## p.pid is also a session ID and a process group ID--we'll use it later to send signals to the whole group
         else:
             ## esto sucede por ejemplo cuando arranca el agente, que estamos en interfaceAdmin() en el mismo hilo, sin _long_running_job ni hilo separado
             #logger.debug ('no queue--not writing any PID to it')
             pass
 
         sout = serr = ''
-        poll_iterations = 1
-        while p.poll() is None:
-            for l in iter (p.stdout.readline, b''):
+        finished = False
+        while True:
+            try:
+                p.wait (0.05)
+                finished = True
+            except subprocess.TimeoutExpired:
+                pass
+
+            ready_to_read, _, _ = select.select ([p.stdout, p.stderr], [], [], 0.2)
+            if p.stdout in ready_to_read:
+                l = p.stdout.readline()
                 partial = l.decode ('utf-8', 'ignore')
                 if self.stdout_q: self.stdout_q.put (partial)
                 sout += partial
-            for l in iter (p.stderr.readline, b''):
+            if p.stderr in ready_to_read:
+                l = p.stderr.readline()
                 partial = l.decode ('utf-8', 'ignore')
                 serr += partial
 
-            ## poll quickly at first, then poll less frequently
-            if   poll_iterations > 15: sleep_time = 1
-            elif poll_iterations > 10: sleep_time = 0.2
-            else:                      sleep_time = 0.1
-            time.sleep (sleep_time)
-            poll_iterations += 1
+            if finished: break
         sout = sout.strip()
         serr = serr.strip()
 

src/stop-agent.ps1

@@ -0,0 +1 @@
+get-process -name ogagentuser|stop-process