tree 455d38cbb0165ab31106ad4a9084d4c48cd25bcd
parent 3641b97cfcc1508404a89b81019819bcb8fd2834
author Bogdan Caprita <caprita@google.com> 1430418104 -0700
committer Bogdan Caprita <caprita@google.com> 1430418106 -0700

profiles/internal/rpc: more information when server hangs waiting on listeners

The most common reason why prod servers fail to shutdown cleanly is that
server.Stop() hangs. This cl makes the following changes to improve our chances
of diagnosing when this happens:

- reduce the timeout after which we print that the server is stuck waiting on
  listeners from 5 mins to 5 secs. 5 minutes is a really long time, and we
  rarely wait that long. It would be good to signal that we're stuck as soon as
  it's likely that we are.

- to mitigate the false positive rate for the above change, print out something
  if the listener wait does in fact complete

- add information about which server is actually stuck (in particular, the type
  of the dispatcher is useful to distinguish servers in a multi-server daemon
  like the mounttable (not to mention that the app cycle manager server is part
  of any server run by device manager))

- remove some unused endpoints field that I found by chance.

Change-Id: I14539c057b9f52a6b9afbc3a18cc0fb455bbbfa5