v1.19.0: "Cannot add fragment of condition ... to parent type"

[marc-barry]

Describe the bug

We rolled out 1.19.0 today and immediately hit an error and found the same error was reported under apollographql/federation#2592. It looks like a PR was created under apollographql/federation#2594 to address the issue.

I wasn't sure how to report you pulled in a dependency from https://github.com/apollographql/federation which broke some queries for specific users. We have since downgraded back to https://github.com/apollographql/router/releases/tag/v1.18.1 while we wait for a new release to pull in a new version of the federation dependency.

To Reproduce

See apollographql/federation#2594 for the fix.

Expected behavior

The behaviour proposed by apollographql/federation#2594.

Output

Error output from what we saw:

{"errors":[{"message":"value retrieval failed: router bridge error: the deno runtime raised an error: `request: couldn't receive response couldn't deserialize payload `11308918137995749901`: `deno: couldn't deserialize response : `Error(\"invalid neither null nor empty object: found Object {\\\"code\\\": String(\\\"QUERY_PLANNING_FAILED\\\"), \\\"exception\\\": Object {\\\"stacktrace\\\": Array [String(\\\"Cannot add fragment of condition \\\\\\\"Entry\\\\\\\" (runtimes: [Entry]) to parent type \\\\\\\"EntryPercent\\\\\\\" (runtimes: PaymentSplitConfigEntryPercent)\\\")]}}\", line: 0, column: 0)``.`","extensions":{"code":"INTERNAL_SERVER_ERROR"}}]}

Desktop (please complete the following information):

• OS: Linux
• Version: v1.19.0

Additional context

N/A

[Meemaw]

Does v1.18.1 work fine for you? For us pretty much every release since 1.15.1 has been broken, due to different serialisation/buffer issues.

[Geal]

@marc-barry thanks for pointing it out. This should be available shortly in the router's query planner
@Meemaw are your serialization or buffer issues related to query planning too? Or is it other issues, like the compression bugs we had?

[Meemaw]

@Geal Seemed more related to compression.

[abernix]

@marc-barry You opened the issue in a fine place! Thank you! In fact, the issue on federation you linked to was previously here on router, but we transferred it to federation to get it fixed by the right team. As you noted, router merely embeds that query-planning component and as @Geal noted, there should be a fix for it soon via a Federation query planner bug fix. Approximations suggest Friday. We'll keep your issue here to track updating the dep on the Router side and close it with the release.

@Meemaw Previous releases you mentioned may have been impacted by different bugs at the time of their release but, but v1.18.1 doesn't exhibit this particular bug as @marc-barry noted in the original post. I've updated the title of this issue to concretely frame this particular bug.

Thanks!

[Geal]

@Meemaw 1.18.1 should have fixed the compression bug with ##3067. If you still see related bugs, please open an issue, I'll take a look shortly

[marc-barry]

Does v1.18.1 work fine for you? For us pretty much every release since 1.15.1 has been broken, due to different serialisation/buffer issues.

Yes. It is working for us and we have currently reverted back to v1.18.1.

[abernix]

We should have a Router release out today with Federation 2.4.6's query planner which fixes this: https://github.com/apollographql/federation/releases/tag/%40apollo%2Fquery-planner%402.4.6.

[abernix]

This is resolved in https://github.com/apollographql/router/releases/tag/v1.19.1.

[marc-barry]

We deployed https://github.com/apollographql/router/releases/tag/v1.19.1 and unfortunately, the router did this for the queries causing issues when we moved to v1.19.0. We have had to revert back to v1.18.1.

{
    "span": {
        "http.flavor": "HTTP/1.1",
        "http.method": "POST",
        "http.route": "/graphql",
        "otel.kind": "SERVER",
        "name": "request"
    },
    "classification": "Status code: 500 Internal Server Error",
    "spans": [
        {
        "http.flavor": "HTTP/1.1",
        "otel.kind": "SERVER",
        "http.method": "POST",
        "http.route": "/graphql",
        "name": "request"
        }
    ],
    "message": "response failed",
    "timestamp": "2023-05-26T16:22:50.211396Z",
    "latency": "0 ms",
    "level": "ERROR"
}

[abernix]

@marc-barry As noted on the other issue, please check again with https://github.com/apollographql/router/releases/tag/v1.20.0.