jwt-proxy doesn't work with SSE (Server Side Events)

[ChristianSchenk711]

I have added and deployed some extra plugins for theia. One of those plugins uses SSE (Server Side Events) events to get notifications. However that did work locally but not after deploying this to our cluster.

I figured out that the problem for me was the jwt-proxy:

SSE works such, that the server writes a few lines to a long-polling HTTP GET whenever it wants to notify the client. However the jwt-proxy will not flush these lines and therefore they will get buffered until the buffer used by io.copy is used is full (which seems to be in the range of 8-16KB which equals to a lot of messages...)

For testing I used a simple SSE Server (like this: https://jasonbutz.info/2018/08/server-sent-events-with-node/) which sents a message every second to subscribers.
If I try to connect to this source behind the jwt-proxy, it will return the messages as bunch, which makes it quite pointless.

I would suggest to handle SSE with an explicit Flush, e.g. like that in proxy.go:

	var finalWriter io.Writer = w
		if w.Header().Get("content-type") == "text/event-stream" {
			finalWriter = NewFlushWriter(w)
		}
		nr, err := io.Copy(finalWriter, resp.Body)

The FlushWriter would just overwrite "write" and call a flush afterwards:

type FlushWriter struct {
	w      io.Writer
	writes uint64
	bytes  uint64
}

func NewFlushWriter(w io.Writer) *FlushWriter {
	return &FlushWriter{w: w}
}

func (w *FlushWriter) Write(b []byte) (int, error) {
	a, err := w.w.Write(b)
	if f, ok := w.w.(http.Flusher); ok {
		f.Flush()
	} else {
		log.Fatal("no flush")
	}
	return a, err
}

With these modifications the simple SSE example properly returns one message per second.

[ChristianSchenk711]

Just realized, that this actually is a problem of the vendored underlying coreos/goproxy in
vendor/github.com/coreos/goproxy/proxy.go, not in proxy/proxy.go So I probably will open a ticket there as well, since that fix needs to go into coreos/goproxy/proxy.go

[ChristianSchenk711]

Reported to coreos/goproxy: coreos/goproxy#5

[che-bot]

Issues go stale after 180 days of inactivity. lifecycle/stale issues rot after an additional 7 days of inactivity and eventually close.

Mark the issue as fresh with /remove-lifecycle stale in a new comment.

If this issue is safe to close now please do so.

Moderators: Add lifecycle/frozen label to avoid stale mode.