Don't write error response when it's already been written

[geoand]

The original failure that Vertx had captured and was trying to be handled by QuarkusErrorHandler was io.netty.channel.StacklessClosedChannelException

Relates to: #17548

[sberyozkin]

Thanks Georgios for a super fast fix, which is also simple, hence even I can approve it :-)

[gsmet]

It might be a dumb question but why do we have an error in the first place given the OP is just clicking around?

[sberyozkin]

@gsmet, Hi Guillaume, the devconsole handlers are implemented as Vert.x handlers, it may explain it

[sberyozkin]

@gsmet see for example here: https://quarkus.io/guides/dev-ui#advanced-usage-adding-actions

[geoand]

aroun

It seems like Netty might close a Channel and cause this.
I don't know if that is new to the latest Netty version or not

[gsmet]

Yeah. For me, that looks very very suspicious. I wouldn't expect Netty to throw random errors after a few simple GETs.

[geoand]

That may or may not be true to be honest. I'll look a little more, but I doubt there is an actual problem here and even if there is, there is likely very little we can do about it

[sberyozkin]

We already had something similar fixed before: https://github.com/quarkusio/quarkus/pull/13060/files

[gsmet]

Well, I'm not saying it would be our own bug to fix, it might be an issue elsewhere. But I fail to see how this could be a normal behavior.

Let's not merge this too fast and wait for @stuartwdouglas to have a look if he can.

[geoand]

It looks like the StacklessClosedChannelException exception is being thrown by io.vertx.core.http.impl.WebSocketImplBase#closeConnection, I guess this is for @pmlopes or @vietj to take a look

[geoand]

Well, I'm not saying it would be our own bug to fix, it might be an issue elsewhere. But I fail to see how this could be a normal behavior.

Let's not merge this too fast and wait for @stuartwdouglas to have a look if he can.

See my latest comment.

It's only happens in web sockets (which the DevUI uses) and in any case this fix should be in place anyway

[sberyozkin]

It's only happens in web sockets (which the DevUI uses) and in any case this fix should be in place anyway

Makes sense, perhaps removing Fixes #17548 would be a good compromise for the underlying cause be investigated as part of #17548 ? preventing the confusing error message in the logs is indeed needed in any case

[geoand]

Sure yeah, done :)

[sberyozkin]

Look forward to rebasing my Keycloak Devservices PR :-),
I suppose we can merge it now, Guillaume, hope you are OK with it, I've added a note to ##17548 that you'd like the lower level issue investigated, Stuart and others may have some ideas

[doublefx]

@sberyozkin

Same issue just happened again using Quarkus 2.1.0-FINAL, opening Swagger

2-Aug 14:04:24.097 INFO  [io.ve.ex.we.RoutingContext] (vert.x-eventloop-thread-7) RoutingContext failure (500): io.netty.channel.StacklessClosedChannelException
	at io.netty.channel.AbstractChannel$AbstractUnsafe.write(Object, ChannelPromise)(Unknown Source)