sarkaramrit1993 commented on PR #5238: URL: https://github.com/apache/solr/pull/5238#issuecomment-6617504289
Thanks David, I've addressed your comments. The one on [`isNestedRequest()`](https://github.com/apache/solr/blob/7f224be7ceca075106804df494468711454da23d/solr/core/src/java/org/apache/solr/request/SolrRequestInfo.java#L70) is answered inline. On tracing: I see the two as complementary, and I want tracing in Solr to get better too. But `always_on` isn't something every production cluster can turn on. At my employer sampling is strictly 1%, and even one server span per request still means exporting and storing every search. Why both: - **Cost decides.** It's the [most common tool-selection criterion in Grafana's 2026 survey](https://grafana.com/observability-survey/2026/), at 65%. A [recent vendor cost guide](https://www.progressiverobot.com/2026/08/11/observability-cost-logs-metrics-traces/) puts typical retention under 5% of trace volume. - **Head sampling is blind to latency.** It [decides when the request starts](https://solr.apache.org/guide/solr/latest/deployment-guide/distributed-tracing.html#sampling). At 1%, 99 of 100 slow searches leave no trace. - **Tail sampling needs a collector tier** buffering whole traces. Elastic [cut that memory by up to 65%, at roughly 2x CPU](https://www.elastic.co/observability-labs/blog/tail-sampling-memory-opentelemetry) (July 2026). The idea is for Solr to get smarter about signalling to tracing which flows need a trace. This hook sees every search and carries its trace id, so it can point at the slow and costly ones. With [SOLR-13732](https://issues.apache.org/jira/browse/SOLR-13732), always tracing a request that carries a configured header, you replay one of those and get its full trace even at 1%. That and CPU time and hit count on the search span, added generically from the response header as you suggested, are the two pieces I took on in [dev@](https://lists.apache.org/thread/t7mft9dbntgtp8pjfh3zvomxh8zdobtd). I'll come back with real use cases after talking to Solr engineers at my employer and elsewhere. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
