Post AXCBFQWg9fTfs8gVM0 by feld@bikeshed.party
(DIR) More posts by feld@bikeshed.party
(DIR) Post #AXCBFPpmj9ANj6cFUm by wolf480pl@mstdn.io
2023-06-29T20:21:50Z
0 likes, 0 repeats
Choosing a time series database for soft-realtime data is frustrating...Prometheus has an excellent query language, but it's exclusively pull-based, which greatly limits the temporal resolution and increases latency.InfluxDB is push-based which means you can have a metric change as many times as you want, at arbitrary timestamps, and you can send those metrics at any time and they'll be immediately received.But Influx's query language is dogshit.
(DIR) Post #AXCBFQWg9fTfs8gVM0 by feld@bikeshed.party
2023-06-29T21:04:23.886872Z
0 likes, 0 repeats
Use telegraf and push into Timescale and you get Postgres query but Influx everything else
(DIR) Post #AXCC75n1T6OdCGv2ES by feld@bikeshed.party
2023-06-29T21:13:58.310781Z
0 likes, 0 repeats
https://github.com/influxdata/telegraf/tree/master/plugins/outputs/postgresql#timescaledbDo this, it's good out of the box. It will automatically create the foreign keys and stuff, you just smash in data and labels into Telegraf via influx or whatever and then Telegraf writes it to timescaleThen you can query it with sanity intact and also generate Timescale's continuous aggregates of data, make views off that so you have a live table of the latest datapoints, etc and it's all real-time and crazy low CPU usagehttps://docs.timescale.com/use-timescale/latest/continuous-aggregates/
(DIR) Post #AXCCAJ3uZv9FEHLpgW by wolf480pl@mstdn.io
2023-06-29T21:09:50Z
0 likes, 0 repeats
@feld hmm timescale is that time series db wrapper around postgres?
(DIR) Post #AXCCAJk634tNL75WRE by feld@bikeshed.party
2023-06-29T21:14:48.119653Z
1 likes, 0 repeats
Yea it's one magical extension you install that will change your life. The benchmarks are insane
(DIR) Post #AXCFU5i6EHsJTQH4a0 by void@bikeshed.party
2023-06-29T21:52:08.694360Z
0 likes, 0 repeats
timescaledb