We are passing all host names of our cluster to the ES reporter. If one of the hosts is not avail. it logs an error, which is misleading, since the next host in the list is up and receiving the data.
I think the log level should be set to warn and only if none of the hosts is available, it should log an error, because only then did we lose data.
Also, it could be useful to randomize the hosts when sending metrics. That way you can easier distribute the load, if many services are reporting into an ES instance.
We are passing all host names of our cluster to the ES reporter. If one of the hosts is not avail. it logs an error, which is misleading, since the next host in the list is up and receiving the data.
I think the log level should be set to warn and only if none of the hosts is available, it should log an error, because only then did we lose data.
Also, it could be useful to randomize the hosts when sending metrics. That way you can easier distribute the load, if many services are reporting into an ES instance.