Niksa Jakovljevic

Add batching for traces

1

Batching will help to achieve better ingest performance, especially if traces are sent one by one. A batch is produced when one of two conditions is met: batch size or...

At the moment we depend on OTEL collector to perform batching of traces. Since we are adding endpoint to support traces coming directly from Jaeger we have to implement batching...

Ease locking to speedup ingestion on startup

1

Due to a lot of locking while inserting metadata on series we tend to have a `slow start`. It can take up to 15-20 minutes to get to full speed...

Running in debug mode produces really low amount of logs thus gives very little visibility into application behaviour. Having more debug logs could be useful for troubleshooting as well.

Consider splitting write pool

1

Currently write pool is shared between traces and metrics. Not many users are ingesting both at the moment, however we should benchmark this and decide if we need to split...

Better defaults for cache sizes

3

To achieve good performance we are currently recommending our users to tweak following: ``` --metrics.cache.series.initial-size=30000000 --cache.memory-target=90% ``` Let's make sure that cache memory defaults work well enough out of the...

Empty resultset on instant query regex

3

**Describe the bug** Promscale returns empty results although data is present in the DB. **To Reproduce** Send query: `/api/v1/query?query=count by (job, instance, version) (prometheus_build_info{job=~".+", instance=~".+"})` Response: `{"status":"success","data":{"resultType":"vector","result":[]}}` **Version** - Distribution/OS:...

Inconsistent connect fallback behaviour for different authentication methods

2

When trying to establish a connection using MD5 auth method, if authentication fails due to wrong password `pgconn` will skip checking fallback configs. However if authentication method used when connecting...

This is useful for CNDB in the context of multi-tenancy.