InfluxDBStore: various improvements (+lower memory usage)

[slimsag]

• Fix an index out of bounds panic when viewing the /traces page.
• Use a less memory intensive DB schema (users will need to rm -rf ~/.influxdb to remove the old DB).
  • In our production application, this change shows a decrease from 40+ GB (and OOM panics) to ~2.4 GB of RAM usage.
• Upgrade to InfluxDB client v2 (fixes #140).

[keegancsmith]

I don't know if I have asked this before, but is influxdb appropriate for the appdash use case? We aren't really storing timeseries data.

But otherwise LGTM

[chris-ramon]

I don't know if I have asked this before, but is influxdb appropriate for the appdash use case? We aren't really storing timeseries data.

Hi @keegancsmith, here an issue where @slimsag wrote about the motivations for replacing AggregateStore for InfluxDBStore you might want to take a look.

Additional to that I'd like to add the following points on why it can be a good use case for appdash:

• Optimized for high-throughput scenarios(Eg. using 1 node, it can ingest a billion values per day).
  • Thousands of spans writes per second.
• Complex queries via it's integrated query language:
  • Better span's info. summary on appdash's dashboard.
• Retention Policies & Continuous Queries:
  • Downsampling & Aggregation of spans.

[keegancsmith]

@chris-ramon cool thanks for the response and comments. Wasn't saying we made the wrong choice, was just looking for motivation :)

The points you mentioned are great, and I'd love to learn more about it. Would you mind a few emails or VC to sate my curiosity on this?

[chris-ramon]

@keegancsmith sounds good :) - I'll be sharing some detailed info. expanding on the points mentioned above.