Hi Bill, From my perspective this is definitely a bug: someone is constructing a syntactically valid but empty response. The correct behavior in the case of an error or timeout would be an HTTP 5xx error, e.g. 503 Service Unavailable or 504 Gateway Timeout.
Inspecting the logs is a good idea, but this is an issue that appears intermittently to people I don't know talking to servers I don't control. If I was able to reproduce this locally with a load test, I could make some progress. Thanks, Ken On Fri, Oct 31, 2025 at 10:39 AM Bill Erickson <[email protected]> wrote: > > Hi Ken, > > The most likely cause of what you are seeing is an error or timeout in the > back-end API that the gateway is proxying. It's probably not the gateway > itself. I'd start by investigating the API calls, check the logs, test via > srfsh, etc. > > -b > > > > On Thu, Oct 30, 2025 at 10:59 AM Ken Cox via Evergreen-dev > <[email protected]> wrote: >> >> Has anyone conducted a load test of the OSRF Gateway? Can you point >> me to code and/or testing artifacts? Left to my own devices I would >> start with Locust (https://locust.io/), because my colleagues have >> used it in the past. >> >> Users of the mobile app sometimes see the error "expected object, got >> empty". This means a request through /osrf-gateway-v1 returned an >> empty payload: >> >> {"payload":[],"status":200} >> >> When I first brought this up >> (https://list.evergreen-ils.org/pipermail/evergreen-dev/2021-January/000083.html), >> I heard "That’s definitely an Evergreen bug" and let it go. But now >> I'm wondering: >> * Could I reproduce this in a load test? >> * Does the Rust gateway have this issue? >> >> Thanks, >> Ken >> _______________________________________________ >> Evergreen-dev mailing list -- [email protected] >> To unsubscribe send an email to [email protected] -- -Ken _______________________________________________ Evergreen-dev mailing list -- [email protected] To unsubscribe send an email to [email protected]
