typesense

Hi team! We see that some of the search request are taking longer than usual, for example &gt;1s. Is it possible to troubleshoot such queries? To not just find them, but also narrow down the reasons why they are slow

In our case it is exactly the same query every minute (query with empty search from our ping check), and in 90% response from typesense instnace takes 200ms and in 10% it takes 1s+

We are measuring latencies inside of the cluster network, so it seems like network is excluded from the equation

Did you also correlate with `search_time_ms` returned by the API?

We set a hard timeout of 1s, so we don’t have access to response for such requests. But it’s a good idea, I’ll start logging reported and real time of requests to find discrepancies. Any other advice for troubleshooting?

You can enable slow search logging via `log-slow-searches-time-ms` server side parameter. Set it to a value like 1000 ms and you will catch all those queries that timeout on the client but complete on the backend (unless you also set a search_cutoff query param.