Stabilitze pov-recovery zombienet-sdk test

[skunert]

closes #8985

I examined the two highlighted failures in the issue. In two cases, node three fails to recover the required blocks.

In the first case, the node is able to recover 16 blocks before encountering:

2025-06-24 06:00:12.022 ERROR ⋮cumulus-relay-chain-streams: [Parachain] Failed to fetch pending candidates. error=JsonRpcError(Call(ErrorObject { code: ServerError(-32000), message: "Error when retrieving the call proof: Call proof query errors:\n- ConnectionShutdown", data: None }))
2025-06-24 06:00:12.023 ERROR ⋮cumulus-relay-chain-streams: [Parachain] Failed to fetch session index. error=JsonRpcError(Call(ErrorObject { code: ServerError(-32000), message: "Error when retrieving the call proof: No node available for call proof query", data: None }))

For the other failure we also see the same RPC-related error. Since for node three, the RPC is the internal light client, I just disabled it and purged it from this test.

We will probably remove support for this experimental feature at some point, but currently does not have high priority. And should for sure not impact the stability of this test.

[paritytech-workflow-stopper]

All GitHub workflows were cancelled due to failure one of the required jobs.
Failed workflow url: https://github.com/paritytech/polkadot-sdk/actions/runs/15921207310
Failed job name: run-frame-omni-bencher

[pepoviola]

Looks great, thanks for the fix @skunert 🙌 . Make sense to add a new issue to track this

We will probably remove support for this experimental feature at some point, but currently does not have high priority. And should for sure not impact the stability of this test.

Thx!

[skunert]

Yep, issue now here #9013

[lrubasze]

LGTM! The simpler the more stable 😁

[lrubasze]

BTW these failing Cumulus CI tests are not related to these changes.
Some regression introduced after merging master yesterday (most likely)

EDIT!
To clarify.
Yesterday I've merged master to ci/zombienet-cumulus-tests-zombienet-sdk.
One of the change was the runner change: parity-large (16 cores) -> parity-zombienet-native-large (8 cores)
And it seems that failing tests are affected with this. Some nodes are not getting enough CPU.

[lrubasze]

additionally full_node_warp_sync fails because of missing method GetParachainInfo_parachain_id

  21   │ 2025-06-27 11:28:24.574  INFO tokio-runtime-worker prometheus: [spiffy-disgust-2437] 〽️ Prometheus exporter started at 127.0.0.1:46145
  22   │ 2025-06-27 11:28:24.574  INFO                 main sc_rpc_server: [spiffy-disgust-2437] Running JSON-RPC server: addr=127.0.0.1:9945,[::1]:9945
  23   │ 2025-06-27 11:28:24.582  INFO                 main cumulus_test_service: [Parachain] Using inprocess node.
  24   │ 2025-06-27 11:28:24.583 ERROR tokio-runtime-worker sc_service::task_manager: [Parachain] Essential task `txpool-background` failed. Shutting down service.
  25   │ 2025-06-27 11:28:24.584 ERROR tokio-runtime-worker sc_service::task_manager: [spiffy-disgust-2437] Essential task `txpool-background` failed. Shutting down service.
  26   │ thread 'main' panicked at cumulus/test/service/src/main.rs:146:18:
  27   │ could not create Cumulus test service: Application(Application(Execution(Other("Exported method GetParachainInfo_parachain_id is not found"))))
  28   │ note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

This is related to this PR #8651

I will disable this test and create an issue to fix it.

[lrubasze]

Failing zombienet-cumulus-0007-full_node_warp_sync issue: #9018

[skunert]

Locally everything looking good, we can merge from my side!