feat: stream bootloader logs

[abuchanan-airbyte]

Stream bootloader logs to the debug output, unless they're at the "ERROR" level, then write them as error logs.

[colesnodgrass]

I wonder how fragile this is. We're likely changing our logging implementation on the platform shortly, and I don't think we have any guarantees that this format won't be modified with that change.

[abuchanan-airbyte]

Yes, all this log stuff is definitely fragile, but it's better than nothing for now. We need to do some work in the bootloader to make all this stuff clean and easy and stable over time.

And, maybe it's reasonable to say that the current log format should be the standard going forward. It's a very common format. Note that there are actually multiple formats in these logs too, which is annoying.

What's best might be to not parse logs in abctl, but have some other summarized bootloader status.

[colesnodgrass]

Can we check the state of the ctx in this loop? Break out if the context is no longer active?

[abuchanan-airbyte]

I don't think that's necessary. The context is passed to the k8s.StreamPodLogs call, which returns a reader, which I would expect to close when the context is canceled.

[abuchanan-airbyte]

I proved this to myself. I couldn't think of a nice way to write an actual test for it (since it depends on the real k8s implementation), but I hacked in this code (note javaLogScanner is not in this PR, but it's just a scanner with the java parsing wrapped in)

func (c *Command) streamPodLogs(ctx context.Context, namespace, podName, prefix string, since time.Time) error {

	ctx, cancel := context.WithCancel(ctx)
	go func() {
		log.Printf("sleeping\n")
		time.Sleep(30 * time.Second)
		log.Printf("canceling\n")
		cancel()
	}()

	r, err := c.k8s.StreamPodLogs(ctx, namespace, podName, since)
	if err != nil {
		return err
	}
	defer r.Close()

	s := newJavaLogScanner(r)
	for s.Scan() {
		if s.line.level == "ERROR" {
			pterm.Error.Printfln("%s: %s", prefix, s.line.msg)
		} else {
			pterm.Debug.Printfln("%s: %s", prefix, s.line.msg)
		}
	}

	log.Printf("scanner exited: %v", s.Err())
	return s.Err()
}

and you can see the scanner exists in the logs: