diff --git a/redash/tasks/general.py b/redash/tasks/general.py
index e0090dfc17..7d2ab0d6bf 100644
--- a/redash/tasks/general.py
+++ b/redash/tasks/general.py
@@ -1,8 +1,7 @@
 import requests
 
 from flask_mail import Message
-from rq import Connection, Queue
-from rq.registry import FailedJobRegistry
+from rq.job import Job
 from redash import mail, models, settings, rq_redis_connection
 from redash.models import users
 from redash.version_check import run_version_check
@@ -65,30 +64,15 @@ def sync_user_details():
 
 
 def purge_failed_jobs():
-    jobs = rq_redis_connection.scan_iter('rq:job:*')
+    jobs = rq_redis_connection.scan_iter('{}*'.format(Job.redis_job_namespace_prefix))
 
     is_idle = lambda key: rq_redis_connection.object('idletime', key) > settings.JOB_DEFAULT_FAILURE_TTL
     has_failed = lambda key: rq_redis_connection.hget(key, 'status') == b'failed'
+    stale_job_keys = [key.decode().split(':').pop() for key in jobs if is_idle(key) and has_failed(key)]
+    stale_jobs = Job.fetch_many(stale_job_keys, rq_redis_connection)
 
-    def not_in_any_failed_registry(key):
-        """ This function should reject any key which is inside any FailedJobRegistry.
-        However, at the moment on RQ v1.1 the @job decorator does not allow setting of a failure_ttl, so jobs
-        are kept inside the FailedJobRegistry for a year and there is no easy way tweak that.
-        This has already been fixed on rq/master (https://github.com/rq/rq/pull/1130) and will be available on the
-        next release. Until then, we simply don't reject keys in a FailedJobRegistry and purge any failed jobs
-        that have been idle for over settings.JOB_DEFAULT_FAILURE_TTL.
-        Once a new RQ release is available, we can delete this comment and the following line:"""
-        return True
-
-        with Connection(rq_redis_connection):
-            failed_registries = [FailedJobRegistry(queue=q) for q in Queue.all()]
-
-        job_id = lambda key: key.decode().split(':').pop()
-        return all([job_id(key) not in registry for registry in failed_registries])
-
-    stale_jobs = [key for key in jobs if is_idle(key) and has_failed(key) and not_in_any_failed_registry(key)]
-
-    for key in stale_jobs:
-        rq_redis_connection.delete(key)
+    for job in stale_jobs:
+        job.delete()
+        rq_redis_connection.delete(job.key)
 
     logger.info('Purged %d old failed jobs.', len(stale_jobs))