Correct hostname not displayed in syslog after boot

I've marked this as affecting cloud-init; I'd be interested to hear smoser's opinion on how he thinks this would best be fixed. (I think he's on vacation this week, so it'll have to wait a few days)

Status changed to 'Confirmed' because the bug affects multiple users.

Can you explain where you see this ?

Just in /var/log.cloud-init.log or elsewhere ?
Its interesting because at the momemnt i'm leaning towards making cloud-init not log through rsyslog unless instructed to.

https://code.launchpad.net/~harlowja/cloud-init/+git/cloud-init/+merge/301729

I originally saw this in /var/log/syslog of all my GCE machines that were built on-top of ubuntu-1604-xenial-v20160610 and older.
See also this bug-report: https://github.com/GoogleCloudPlatform/compute-image-packages/issues/290

I'm now running machines off the image ubuntu-1604-xenial-v20160907a where I can't reproduce the problem any more.

OK, maybe I was too quick on this one.
Just did a test with ubuntu-1604-xenial-v20161013 in GCE.

To reproduce, simply boot up a machine from that image and look at the /var/log/syslog.
The hostname in the log says "ubuntu" instead of the real machine hostname.

Sending a testlog with "logger testlog1" gives me this line in /var/log/syslog:
Oct 18 08:20:22 ubuntu igottwald: testlog 1

After that I do a "sudo pkill -HUP rsyslogd" and write another testlog with "logger testlog 2", then I see in /var/log/syslog:
Oct 18 08:26:32 test-ig1 igottwald: testlog 2

test-ig1 is the correct hostname.

Same is true if you build images with packer on-top of this image. The resulting images start with a hostname in the log which looks like "packer-bc6a04da-56ea-4982-aed6-664877993834" since this is the method of packer choosing random hostnames during image build.
Again once the rsyslogd is HUPed or restarted, the correct hostname is back again.