no more index entries in atom_tab (max=1048576)

[huangyonghome]

Hi
we often encounter below problem on several Rabbitmq cluster nodes since we have installed this monitor script:

Oct  7 04:11:25 mq-es-node3 rabbitmq-server[2274]: Crash dump is being written to: erl_crash.dump...done  
Oct  7 04:11:25 mq-es-node3 rabbitmq-server[2274]: no more index entries in atom_tab (max=1048576)  
Oct  7 04:11:25 mq-es-node3 systemd[1]: rabbitmq-server.service: Main process exited, code=exited, status=1/FAILURE  
Oct  7 04:11:26 mq-es-node3 rabbitmqctl[2511]: Stopping and halting node 'rabbit@mq-es-node3' ...  
Oct  7 04:11:26 mq-es-node3 rabbitmqctl[2511]: Error: unable to connect to node 'rabbit@mq-es-node3': nodedown  

I digged this issue on Google.and found the same problem on Github Rabbimt project.I saw this words:

This issue is marked as fixed, but Rabbit 3.6.5 still runs out of memory with the same symptom as 3.6.3 when deployed to PCF,
and we were told it was due to atom table exhaustion. 
We even put in a config param to increase atom table size, which helped. 
What's unique about PCF, is that it runs a metrics collector task that periodically hits Rabbit management API.

So,I am wondering if there are some behaviors of the zabbix monitoring script to raise the atom ?

rabbitmq version: 3.6.3
erl version: 18.3

Could you please help me to check on this ? Thank you in advance

[jasonmcintosh]

This sounds like a bug in rabbitmq's management plugin. Unfortunately, I don't do much with rabbitmq/zabbix anymore, so I'll be unable to assist :( At this point, most of the monitoring I've seen is heading towards Prometheus or similar and I've been heading more that direction.

[huangyonghome]

@jasonmcintosh Thank you for your explanation . I tend to consider it as a BUG of rabbitmq itself