Routing by origin - service is not good enough

[dmedinag]

Component(s)

exporter/loadbalancing

Is your feature request related to a problem? Please describe.

There exist two keys for routing spans today: service and traceID. traceID is great for tail sampling, while service is the only one producing reliable spanmetrics without insane cardinality on big volumes.

Routing spans by service, however, brings a big problem to the table if the different services generating spans produce a very uneven volume. In an environment where a small percentage of services produce >99% of the spans, instances of the otelcol running the spanmetrics connector (or any other trace pipeline receiving spans routed by service) sustain a very asymmetric load, overloading a few instances while others stay idle.

Describe the solution you'd like

service does a good job for spanmetrics because it produces metric series issued by a single collector of the layer behind the load balancing collector, easing up aggregation and range functions requiring multiple datapoints. But for this reason so would pod_name in a k8s environment, or any other proxy for task instance (where task = application that generates spans).

In the context of kubernetes, the IP of the incoming connection uniquely identifies a task instance. My proposal is to add source_ip as a routing key for signals, which would considerably dilute the problem we have with service as routing key by seggregating the load of these big services across as many workers as instances that service has. Since big services are also typically the ones with the most instances in our ecosystem (and others in the industry), while it's not as good as traceID in terms of randomness this is a much more efficient routing key and an effective solution for our problem.

Note 1: this solution would also solve this issue to and any other environment where the IP uniquely identifies a task instance. Note 2: this routing key would be applicable for spans, but also for metrics and logs.

Describe alternatives you've considered

Ad-hoc pipeline for offenders

Again within a kubernetes environment, we've considered omitting the span traffic of these big services on our central spanmetrics pipeline and deploying otelcols as sidecars with such spanmetrics pipeline.

But this really just hacks the problem, increases our architecture complexity and does not solve the real issue of service being an inherently bad key for symmetric load.

Hack the routing key

Since we have the ability to modify service before it's used as routing key, we introduced a transformer that changes the value of service to {serviceName}%{ip}, to force the loadbalancer exporter to route by this composed key. Then, on the second layer of collectors, we recover the original value of service by removing the artificial suffix and proceed with the processing. This has mitigated the problem for us for now, we see much more symmetric load in our spanmetrics collectors, but again it's just hacking around a limitation on the exporter.

This solution was preferred to us over having a custom adaptation of the loadbalancingexporter on our side so that we simplify the process of maintaining our collectors.

Additional context

No response

[github-actions[bot]]

Pinging code owners:

• exporter/loadbalancing: @jpkrohling

  See Adding Labels via Comments if you do not have permissions to add labels yourself.