Is it required to restart the routing/coordinator process when data node disconnects from network and comes back?


#1

I know this could have been discussed in several posts. But I couldn't find a solution yet. This happens in few cases[not always]. When a data node moves out of the network and comes back, it is not getting registered to cluster automatically. I end up restarting the routing node process to identify the data node

Is it required to restart the routing/coordinator process when data node disconnects from network and comes back?
Or is there any setting I can change to increase the number of checks from cluster for new nodes when a node goes out of network?


(Mark Walkom) #2

It shouldn't need a restart, no.
Assuming you are using zen discovery, take a look at https://www.elastic.co/guide/en/elasticsearch/reference/2.2/modules-discovery-zen.html


#3

Thanks for the response. I use zen discovery. I only use the zen minimum_master_nodes settings in the config.

zen:
minimum_master_nodes: 2

Do I need set the ping_retries[default to 3] setting if the network failure exists for more than the default retries time 90s. [3 * ping_timeout = 3 * 30s = 90s]

I mean set ping_retries to a higher value [eg. 10 or 20].
So the node will join back in case if the network failure exists for 10 mins with
ping_retries : 20
ping_timeout:30s


(system) #4