Fix CI

[dandavison]

No description provided.

[dandavison]

@Sushisource @roxblnfk LGTY?

[dandavison]

This is addressing the new server change to update errors: when a workflow completes and an update is in Accepted state, that used to be gRPC NOT FOUND, but is now an update failure. All SDKs are having to address this change today, because temporal CLI recently was updated to the new server version. Without the change here, we are getting

AcceptedUpdateCompletedWorkflow: Workflow Update failed because the Workflow completed before the Update completed.

I have to admit I don't fully understand where the gRPC NOT FOUND was being handled, but the fix here seems like good practice and fixes the test.

[Sushisource]

Yeah I think this change is fine for the purposes of this PR.

[roxblnfk]

I think we can just suppress this exception: temporalio/sdk-php@0f0ceec

[dandavison]

Thanks @roxblnfk. So I guess both work -- I think I'm finding the wait allHandlersFinished a bit cleaner since it's something that we recommend. WDYT?

[roxblnfk]

What about $this->counter >= 2 && Workflow::allHandlersFinished()?
Looks like that counter may be important

Yeah. Looks good for me.

[roxblnfk]

@dandavison
There should be a condition $this->counter >= 2 && Workflow::allHandlersFinished() because the Workflow might start and finish before the Update methods are called.

[dandavison]

There are two remaining problems here:

  docker run --rm -i -v /tmp/temporal-certs:/certs features:php-2.12.3 \
  --server $TEMPORAL_CLOUD_ADDRESS \
  --namespace $TEMPORAL_CLOUD_NAMESPACE \
  --client-cert-path /certs/client.pem \
  --client-key-path /certs/client.key
  shell: /usr/bin/bash -e {0}
  env:
    TEMPORAL_CLOUD_ADDRESS: sdk-ci.a2dd6.tmprl.cloud:7233
    TEMPORAL_CLOUD_NAMESPACE: sdk-ci.a2dd6
    REPO_URL: https://github.com/temporalio/features
    FEATURES_BUILT_IMAGE_TAG: features:php-2.12.3
    FEATURES_BUILT_IMAGE_TAGS: features:php-2.12.3;features:php-2.12;features:php-2;features:php
Starting RoadRunner... error
Error starting RoadRunner: handle_serve_command: Function call error:
	serve error from the plugin *rrtemporal.Plugin stopping execution, error: temporal_plugin_serve: failed reaching server: last connection error: connection error: desc = "error reading server preface: read tcp 172.17.0.2:53200->44.241.31.151:7233: read: connection reset by peer"

and

  docker run --rm -i -v /tmp/temporal-certs:/certs features:go-1.32.1 \
  --server $TEMPORAL_CLOUD_ADDRESS \
  --namespace $TEMPORAL_CLOUD_NAMESPACE \
  --client-cert-path /certs/client.pem \
  --client-key-path /certs/client.key

panic: 
	Error Trace:	/app/features/update/updateutil/history.go:20
	            				/app/features/update/activities/feature.go:49
	            				/app/harness/go/harness/runner.go:120
	            				/app/harness/go/cmd/run.go:271
	            				/app/harness/go/cmd/run.go:230
	            				/app/harness/go/cmd/run.go:246
	            				/app/harness/go/cmd/run.go:38
	            				/go/pkg/mod/github.com/urfave/cli/v2@v2.3.0/command.go:163
	            				/go/pkg/mod/github.com/urfave/cli/v2@v2.3.0/app.go:313
	            				/go/pkg/mod/github.com/urfave/cli/v2@v2.3.0/app.go:224
	            				/app/harness/go/cmd/cmd.go:20
	            				/app/prepared/main.go:9
	            				/usr/local/go/src/runtime/proc.go:267
	            				/usr/local/go/src/runtime/asm_amd64.s:1650
	Error:      	Received unexpected error:
	            	after 5s, no workflow(s) found

[rustatian]

Hey @dandavison 👋🏻
RR can't connect to the temporal, looks like either address in the .rr.yaml is not correct or RR (by the provided address) don't have an access to the temporal. RR uses SDK-GO under the hood.

[rustatian]

I double checked the configuration (.rr.yaml) and the problem is in TLS I suppose. RR uses the regular, non-encrypted connection, but you're trying to connect using TLS.
You need to update RR's configuration to include these certs (or envs pointing to these certs) as shown here: link

[dandavison]

Thanks @rustatian! Do you know what recent change might have caused this to break?

[rustatian]

Hard to say, might be some update to the Temporal cloud (or instance located by this address 44.241.31.151:7233) which is now requires a TLS connection only... As far as I know, this is not a SDK-GO update.

[roxblnfk]

You can uncomment this line to check all the command line parameters

features/harness/php/src/Runtime/Runner.php

Line 48 in 5fa5ba7

// echo "\e[1;36mStart RoadRunner with command:\e[0m {$command}\n";

---

The reason might be that the temporal. prefix is missing there:

features/harness/php/src/Runtime/Runner.php

Lines 44 to 45 in 5fa5ba7

	$run->tlsKey === null or $rrCommand = [...$rrCommand, '-o', "tls.key={$run->tlsKey}"];
	$run->tlsCert === null or $rrCommand = [...$rrCommand, '-o', "tls.cert={$run->tlsCert}"];

temporal.tls.key, temporal.tls.cert