What will happen if logs are generated faster than loggly can consume them?

[aomega]

If the service generates a large number of logs over a particular period, and loggly is down or simply consuming them more slowly than they are being generated, is there a mechanism that will stop memory from overflowing in the service? For example, by discarding older or newer logs if the number or size of the pending logs goes over a certain threshold?

[indexzero]

You will lose logs. That's the whole idea behind winston. You should be pushing your logs to redis or something like that as well.