Error with Redis / Distributed

Help
1

====================================

Component Version
LibreNMS 1.60-81-gcf023b9
DB Schema 2020_02_10_223323_create_alert_location_map_table (159)
PHP 7.2.28-3+ubuntu16.04.1+deb.sury.org+1
MySQL 10.0.38-MariaDB-0ubuntu0.16.04.1
RRDTool 1.5.5
SNMP NET-SNMP 5.7.3

====================================

Switched to distributed poller.
Added memcached, Redis and rrdcached.
It was all working until I restarted the poller service on the old “master” instance.
I got an infinite number of

Poller_0-6(ERROR):Poller poller exception! 'Redis' object has no attribute 'bzpopmin'
AttributeError: 'Redis' object has no attribute 'bzpopmin'
Traceback (most recent call last):
  File "/opt/librenms/LibreNMS/queuemanager.py", line 54, in _service_worker
    device_id = self.get_queue(queue_id).get(True, 10)
  File "/opt/librenms/LibreNMS/__init__.py", line 323, in get
    item = self._redis.bzpopmin(self.key, timeout=timeout)
AttributeError: 'Redis' object has no attribute 'bzpopmin'
Poller_0-12(ERROR):Poller poller exception! 'Redis' object has no attribute 'bzpopmin'
AttributeError: 'Redis' object has no attribute 'bzpopmin'
Poller_0-5(ERROR):Poller poller exception! 'Redis' object has no attribute 'bzpopmin'
Traceback (most recent call last):
  File "/opt/librenms/LibreNMS/queuemanager.py", line 54, in _service_worker
    device_id = self.get_queue(queue_id).get(True, 10)
  File "/opt/librenms/LibreNMS/__init__.py", line 323, in get
    item = self._redis.bzpopmin(self.key, timeout=timeout)
AttributeError: 'Redis' object has no attribute 'bzpopmin'

And no device is polled, of course.
Redis is on the same host, configured .env as
REDIS_HOST=127.0.0.1
REDIS_PORT=6379
REDIS_DB=0

I can access the redis-cli and issue commands. Redis is 5.0.7.
This is the node that was running since months, the master and unique node.
The strange thing is that the poller node (with 2 devices) seems to work fine, connecting to mysql/redis/memcached on this same host.
I tried disabling distributed polling on main node in config but I don’t know why it requires this Redis thing now. I’m loosing important data and don’t mind to loose the distributed polling right now, but no way to go back to old config.
Please is there someone who can tell me how to disable the Redis requirement or, better, why it complains about a redis command that can be issued with no problem from the console?
Thanks

2

Is it normal that both nodes are listed in the web interface as cluster master=YES?
image

3

Definitely strange: I stopped Redis on the master node, the one that gives the python exceptions above.
Restarted the poller service: it started. It seems to check if redis is on localhost, even with .env REDIS_X lines removed. If the redis server is available, it starts giving errors. If there is no Redis listening it runs fine.
After starting the librenms poller service, I started the Redis instance and the remote poller works.
So the problem is the same as above: starting redis after poller is a trick, why the poller gives all those errors if Redis is available, while the remote node doesn’t?
And where does the poller service gets its Redis configuration, as I currently have no REDIS_x lines in my .env file?