Persist Telegraf Metric Buffer to filesystem

[prashanthjbabu]

Feature Request

Opening a feature request kicks off a discussion.

Proposal:

It would be great if we could have the option to persist the metric buffer of telegraf to filesystem instead of memory ( this could be an additional option like a flag or a config) .

Current behavior:

Telegraf only stores the metric buffer in memory

Desired behavior:

Store telegraf metric buffer in filesystem

Use case:

This feature would be highly useful for devices which run telegraf which could lose power or crash (and hence reboot) for whatever reason . This will help us to get some vital metrics about the system which would otherwise be lost if stored only in memory. Once the devices comes back up , it can fetch the buffered metrics from file and send it back to the cloud.

[voiprodrigo]

You can use an intermediate local store for this. Route all metrics from your regular inputs to the local store output (such as NSQ, for example, which can be configured to use file storage). Then, in parallel, also have telegraf read from the local store and push to your InfluxDB or whatever destination. The difference here is that the store is not internal to Telegraf, so you need a separate service running on the system, and you need to use the metric route capabilities of Telegraf. Haven’t done it in practice, but in theory should work.

_{Sent with GitHawk}

[PWSys]

+1 This would be great.

[danielnelson]

Duplicate #802