[IMP] Multi-node high-availability jobrunner

queue_job/README.rst

@@ -132,6 +132,23 @@ Configuration
 .. [1] It works with the threaded Odoo server too, although this way
        of running Odoo is obviously not for production purposes.
 
+* Deploying in high availability mode or odoo.sh:
+
+When deploying queue_job on multiple nodes or on odoo.sh, on top of the configuration
+parameters mentioned above you need to also set the env variable
+ODOO_QUEUE_JOB_HIGH_AVAILABILITY=1 or via config parameter as such:
+
+.. code-block:: ini
+
+  (...)
+  [queue_job]
+  high_availability = 1
+
+
+> :warning: **Warning:** Failure to enable the high_availability flag on odoo.sh could
+constitute a breach of Acceptable Use Policy. Always enable this flag via the odoo.conf file for odoo.sh
+deployment
+
 Usage
 =====
 
@@ -571,6 +588,12 @@ Known issues / Roadmap
   You must therefore requeue them manually, either from the Jobs view,
   or by running the following SQL statement *before starting Odoo*:
 
+* When deployed in high_availability mode the allocated databases for the
+  jobrunners must be identical. If the databases are different and overlap
+  i.e jobrunner A runs on DB1,DB2 and jobrunner B runs on DB2,DB3 then either
+  DB1 or DB3 will not proccess jobs because there can be only one leader per 
+  sets of databases.
+
 .. code-block:: sql
 
   update queue_job set state='pending' where state in ('started', 'enqueued')
@@ -631,6 +654,8 @@ Contributors
 * Souheil Bejaoui <[email protected]>
 * Eric Antones <[email protected]>
 * Simone Orsi <[email protected]>
+* Paul Catinean <[email protected]>
+* Ruchir Shukla <[email protected]>
 
 Maintainers
 ~~~~~~~~~~~

queue_job/jobrunner/runner.py

@@ -145,6 +145,7 @@
 import selectors
 import threading
 import time
+import uuid
 from contextlib import closing, contextmanager
 
 import psycopg2
@@ -159,6 +160,7 @@
 
 SELECT_TIMEOUT = 60
 ERROR_RECOVERY_DELAY = 5
+LEADER_CHECK_DELAY = 10
 
 _logger = logging.getLogger(__name__)
 
@@ -192,7 +194,7 @@ def _odoo_now():
     return _datetime_to_epoch(dt)
 
 
-def _connection_info_for(db_name):
+def _connection_info_for(db_name, jobrunner_ha_uuid=None):
     db_or_uri, connection_info = odoo.sql_db.connection_info_for(db_name)
 
     for p in ("host", "port", "user", "password"):
@@ -202,6 +204,8 @@ def _connection_info_for(db_name):
 
         if cfg:
             connection_info[p] = cfg
+        if jobrunner_ha_uuid:
+            connection_info["application_name"] = "jobrunner_%s" % jobrunner_ha_uuid
 
     return connection_info
 
@@ -260,14 +264,13 @@ def urlopen():
 
 
 class Database(object):
-    def __init__(self, db_name):
+    def __init__(self, db_name, jobrunner_ha_uuid=None):
         self.db_name = db_name
-        connection_info = _connection_info_for(db_name)
+        self.jobrunner_ha_uuid = jobrunner_ha_uuid
+        connection_info = _connection_info_for(db_name, self.jobrunner_ha_uuid)
         self.conn = psycopg2.connect(**connection_info)
         self.conn.set_isolation_level(ISOLATION_LEVEL_AUTOCOMMIT)
         self.has_queue_job = self._has_queue_job()
-        if self.has_queue_job:
-            self._initialize()
 
     def close(self):
         # pylint: disable=except-pass
@@ -280,6 +283,41 @@ def close(self):
             pass
         self.conn = None
 
+    def _check_leader(self, jobrunner_db_names):
+        """Check if the linked jobrunner is the leader of all jobrunner_db_names"""
+        if not self.jobrunner_ha_uuid:
+            return False
+
+        with closing(self.conn.cursor()) as cr:
+            cr.execute(
+                """
+                SELECT substring(application_name FROM 'jobrunner_(.*)')
+                FROM pg_stat_activity
+                WHERE application_name LIKE 'jobrunner_%%' AND
+                datname IN %s
+                ORDER BY backend_start, datname
+                LIMIT 1;""",
+                (jobrunner_db_names,),
+            )
+            res = cr.fetchone()
+            leader_uuid = res[0] if res else ""
+            if leader_uuid != self.jobrunner_ha_uuid:
+                _logger.debug(
+                    "jobrunner %s: not leader of db(s) [ %s ]. leader: %s. sleeping %s sec.",
+                    self.jobrunner_ha_uuid,
+                    ", ".join(jobrunner_db_names),
+                    leader_uuid,
+                    LEADER_CHECK_DELAY,
+                )
+                return False
+
+            _logger.info(
+                "jobrunner %s is now the leader of db(s) [ %s ]",
+                self.jobrunner_ha_uuid,
+                ", ".join(jobrunner_db_names),
+            )
+            return True
+
     def _has_queue_job(self):
         with closing(self.conn.cursor()) as cr:
             cr.execute(
@@ -353,6 +391,7 @@ def __init__(
         user=None,
         password=None,
         channel_config_string=None,
+        high_availability=None,
     ):
         self.scheme = scheme
         self.host = host
@@ -363,6 +402,10 @@ def __init__(
         if channel_config_string is None:
             channel_config_string = _channels()
         self.channel_manager.simple_configure(channel_config_string)
+        self.uuid = False
+        if high_availability:
+            self.uuid = str(uuid.uuid4())
+            _logger.info("jobrunner %s initialized in HA mode", self.uuid)
         self.db_by_name = {}
         self._stop = False
         self._stop_pipe = os.pipe()
@@ -388,12 +431,18 @@ def from_environ_or_config(cls):
         password = os.environ.get(
             "ODOO_QUEUE_JOB_HTTP_AUTH_PASSWORD"
         ) or queue_job_config.get("http_auth_password")
+        if "ODOO_QUEUE_JOB_HIGH_AVAILABILITY" in os.environ:
+            high_availability = str(os.environ["ODOO_QUEUE_JOB_HIGH_AVAILABILITY"])
+        else:
+            high_availability = str(queue_job_config.get("high_availability"))
+        high_availability = high_availability.lower() in ("true", "1", "t")
         runner = cls(
             scheme=scheme or "http",
             host=host or "localhost",
             port=port or 8069,
             user=user,
             password=password,
+            high_availability=high_availability,
         )
         return runner
 
@@ -421,15 +470,20 @@ def close_databases(self, remove_jobs=True):
                 _logger.warning("error closing database %s", db_name, exc_info=True)
         self.db_by_name = {}
 
+    def initialize_runner(self):
+        """Listen for db notifications and load existing jobs into memory"""
+        for db in self.db_by_name.values():
+            db._initialize()
+            with db.select_jobs("state in %s", (NOT_DONE,)) as cr:
+                for job_data in cr:
+                    self.channel_manager.notify(db.db_name, *job_data)
+            _logger.info("queue job runner ready for db %s", db.db_name)
+
     def initialize_databases(self):
         for db_name in self.get_db_names():
-            db = Database(db_name)
+            db = Database(db_name, self.uuid)
             if db.has_queue_job:
                 self.db_by_name[db_name] = db
-                with db.select_jobs("state in %s", (NOT_DONE,)) as cr:
-                    for job_data in cr:
-                        self.channel_manager.notify(db_name, *job_data)
-                _logger.info("queue job runner ready for db %s", db_name)
 
     def run_jobs(self):
         now = _odoo_now()
@@ -511,6 +565,17 @@ def stop(self):
         # wakeup the select() in wait_notification
         os.write(self._stop_pipe[1], b".")
 
+    def check_db_leader(self):
+        """Check if the current jobrunner is the leader for all configured databases"""
+        jobrunner_db_names = tuple(self.db_by_name.keys())
+
+        if not jobrunner_db_names:
+            return False
+
+        # Use the first db connection to for leadership check
+        db_obj = self.db_by_name[jobrunner_db_names[0]]
+        return db_obj._check_leader(jobrunner_db_names)
+
     def run(self):
         _logger.info("starting")
         while not self._stop:
@@ -520,6 +585,13 @@ def run(self):
                 # TODO: how to detect new databases or databases
                 #       on which queue_job is installed after server start?
                 self.initialize_databases()
+                while not self._stop and self.uuid:
+                    leader = self.check_db_leader()
+                    if leader:
+                        break
+                    time.sleep(LEADER_CHECK_DELAY)
+                    continue
+                self.initialize_runner()
                 _logger.info("database connections ready")
                 # inner loop does the normal processing
                 while not self._stop:

queue_job/readme/CONFIGURE.rst

@@ -41,3 +41,20 @@
 
 .. [1] It works with the threaded Odoo server too, although this way
        of running Odoo is obviously not for production purposes.
+
+* Deploying in high availability mode or odoo.sh:
+
+When deploying queue_job on multiple nodes or on odoo.sh, on top of the configuration
+parameters mentioned above you need to also set the env variable
+ODOO_QUEUE_JOB_HIGH_AVAILABILITY=1 or via config parameter as such:
+
+.. code-block:: ini
+
+  (...)
+  [queue_job]
+  high_availability = 1
+
+
+> :warning: **Warning:** Failure to enable the high_availability flag on odoo.sh could
+constitute a breach of Acceptable Use Policy. Always enable this flag via the odoo.conf file for odoo.sh
+deployment

queue_job/readme/CONTRIBUTORS.rst

@@ -10,3 +10,5 @@
 * Souheil Bejaoui <[email protected]>
 * Eric Antones <[email protected]>
 * Simone Orsi <[email protected]>
+* Paul Catinean <[email protected]>
+* Ruchir Shukla <[email protected]>

queue_job/readme/ROADMAP.rst

@@ -13,6 +13,12 @@
   You must therefore requeue them manually, either from the Jobs view,
   or by running the following SQL statement *before starting Odoo*:
 
+* When deployed in high_availability mode the allocated databases for the
+  jobrunners must be identical. If the databases are different and overlap
+  i.e jobrunner A runs on DB1,DB2 and jobrunner B runs on DB2,DB3 then either
+  DB1 or DB3 will not proccess jobs because there can be only one leader per 
+  sets of databases.
+
 .. code-block:: sql
 
   update queue_job set state='pending' where state in ('started', 'enqueued')

queue_job/static/description/index.html

@@ -492,6 +492,20 @@ <h1><a class="toc-backref" href="#toc-entry-2">Configuration</a></h1>
 of running Odoo is obviously not for production purposes.</td></tr>
 </tbody>
 </table>
+<ul class="simple">
+<li>Deploying in high availability mode or odoo.sh:</li>
+</ul>
+<p>When deploying queue_job on multiple nodes or on odoo.sh, on top of the configuration
+parameters mentioned above you need to also set the env variable
+ODOO_QUEUE_JOB_HIGH_AVAILABILITY=1 or via config parameter as such:</p>
+<pre class="code ini literal-block">
+<span class="na">(...)</span><span class="w">
+</span><span class="k">[queue_job]</span><span class="w">
+</span><span class="na">high_availability</span><span class="w"> </span><span class="o">=</span><span class="w"> </span><span class="s">1</span>
+</pre>
+<p>&gt; :warning: <strong>Warning:</strong> Failure to enable the high_availability flag on odoo.sh could
+constitute a breach of Acceptable Use Policy. Always enable this flag via the odoo.conf file for odoo.sh
+deployment</p>
 </div>
 <div class="section" id="usage">
 <h1><a class="toc-backref" href="#toc-entry-3">Usage</a></h1>
@@ -871,6 +885,11 @@ <h1><a class="toc-backref" href="#toc-entry-11">Known issues / Roadmap</a></h1>
 therefore fill the running queue and prevent other jobs to start.
 You must therefore requeue them manually, either from the Jobs view,
 or by running the following SQL statement <em>before starting Odoo</em>:</li>
+<li>When deployed in high_availability mode the allocated databases for the
+jobrunners must be identical. If the databases are different and overlap
+i.e jobrunner A runs on DB1,DB2 and jobrunner B runs on DB2,DB3 then either
+DB1 or DB3 will not proccess jobs because there can be only one leader per
+sets of databases.</li>
 </ul>
 <pre class="code sql literal-block">
 <span class="k">update</span><span class="w"> </span><span class="n">queue_job</span><span class="w"> </span><span class="k">set</span><span class="w"> </span><span class="k">state</span><span class="o">=</span><span class="s1">'pending'</span><span class="w"> </span><span class="k">where</span><span class="w"> </span><span class="k">state</span><span class="w"> </span><span class="k">in</span><span class="w"> </span><span class="p">(</span><span class="s1">'started'</span><span class="p">,</span><span class="w"> </span><span class="s1">'enqueued'</span><span class="p">)</span>
@@ -930,6 +949,8 @@ <h2><a class="toc-backref" href="#toc-entry-17">Contributors</a></h2>
 <li>Souheil Bejaoui &lt;<a class="reference external" href="mailto:souheil.bejaoui&#64;acsone.eu">souheil.bejaoui&#64;acsone.eu</a>&gt;</li>
 <li>Eric Antones &lt;<a class="reference external" href="mailto:eantones&#64;nuobit.com">eantones&#64;nuobit.com</a>&gt;</li>
 <li>Simone Orsi &lt;<a class="reference external" href="mailto:simone.orsi&#64;camptocamp.com">simone.orsi&#64;camptocamp.com</a>&gt;</li>
+<li>Paul Catinean &lt;<a class="reference external" href="mailto:pca&#64;pledra.com">pca&#64;pledra.com</a>&gt;</li>
+<li>Ruchir Shukla &lt;<a class="reference external" href="mailto:ruchir&#64;bizzappdev.com">ruchir&#64;bizzappdev.com</a>&gt;</li>
 </ul>
 </div>
 <div class="section" id="maintainers">