commit e8b0213e57e76c908527948740304656e6268cb4
author Adam Sadovsky <sadovsky@google.com> Tue Dec 08 16:36:32 2015 -0800
committer Adam Sadovsky <sadovsky@google.com> Wed Dec 09 00:51:21 2015 +0000
tree 8dc1445bf38a09e6c7010c7181f7a6e790d9aace
parent 603c1b28c2a95bb65b14b6ff80398065b020f080

TBR syncbase: increase healthInfoTimeOut

This fixes a peer selection slowdown problem introduced by
v.io/c/18414:
- v.io/c/18414 was needed because a 200ms connection timeout
  caused sync to never succeed for Kishore's GCE -> proxied
  phone configuration.
- This CL is needed because if healthInfoTimeOut is less
  than connectionTimeOut, we can end up in a similar place
  to where we started, where an unresponsive peer causes
  sync to slow down.

With this CL, we expect the following behavior: when a
device becomes unresponsive, its peers may have up to a 4s
"hiccup" in sync. But after this hiccup, sync should again
be fast. This isn't ideal, and we know of various ways to
improve things further, but it solves the immediate problem.

Ali patched this CL to test it and confirmed that the new
behavior is as described above.

Change-Id: I5e8f8dd82876fe121e7cd00c11b1157143bf719d