site_graphlogo
  -   Terms of Use and Privacy
Demos | Example Code | Meta | Operation
rss
site_graphlogo
  -   Terms of Use and Privacy
Demos | Example Code | Meta | Operation
rss

<<   <   >   >>

2021-07-02 | Meta | Telegraf -> TimescaleDB

I'm still struggling with the best way to unify time-series data and graph data, from the roots on up to leaves, so to speak, with infinite tree/turtle silhouettes on the line, lol. After looking at almost every collector, middleware connector and database, I decided that Telegraf -> TimescaleDB was what I was looking for. I'm still testing. It is possible that I'll use Telegraf -> Fluentd -> TimescaleDB, but every extra connector has consequences. Another possibility that might be worth the effort is CollectD -> Fluent-bit -> Lua -> Fluent-bit -> Postgres (which TimescaleDB runs on), which has the advantage of coming from a straightforward ecosystem (read on).

Back to Telegraf -> TimescaleDB. This is simple from a high level perspective. For years, this connector has been waiting to be merged into the main Telegraf branch. It seems like it is close, so I decided to recompile using Patrick Hemmer's unmerged branch that has his Postgres output changes. I had to install go on my machine, which is running my own OS from kernel through to the analysis tools. I'm doing this, because the only way to truly know if you own your IT is if you can take a DVD/Blu-ray of data, unconnected to the internet, and create your system from a stream of data... here we are back at time-series and why. Here is the list of packages needed to compile. (!!!)

triple virtuoso timescaledb collectd