When a Zookeeper Leader node goes inactive, will the Management server stop until a new ZK leader is identified?

[#1]

In performing a increase of CPU/Memory for one of the Cassandra/Zookeeper nodes, we noticed one of our management server services stopped. We identifed the CA/ZK node being worked was identified as the leader. Knowing the ZK voters would identify a new leader for the system, we also noticed the management server services did not automatically restart. We monitored for a short period of time, then manually performed the all-start.sh to bring up the management services.

Our question, should the management services started after a new ZK leader was identified? As we have 2 management servers on our system, what would cause the management services to stop on one of 2 management servers?

[#2 · Accepted solution]

Hi there!

I'm not sure how you ran into this issue. As far as the auto restarts are concerned, you can configure all the components to perform autostart when system goes down/rebooted, etc. Here is the command to configure it:

/opt/apigee/apigee-service/bin/apigee-all enable_autostart

Hope this helps.

Thanks,

Archendra

[#3]

Thanks @archendra, autostart is an option we've explored.

The ZK under work was the leader. However, when we do upgrades, we have not experienced this situation before with all our upgrade and infrastructure maintenance work. That's the mystery.