]> git.michaelhowe.org Git - packages/o/openafs.git/commit
Rx: Treat rx_minPeerTimeout not as a minimum but as padding
authorJeffrey Altman <jaltman@your-file-system.com>
Thu, 21 Oct 2010 18:13:03 +0000 (14:13 -0400)
committerDerrick Brashear <shadow@dementia.org>
Tue, 26 Oct 2010 02:23:10 +0000 (19:23 -0700)
commite339f372a22d1d4586539bacdf60cf8bc8f0ec47
tree0fa53dfb274ce83c1e50f1c0dd32933c3c63dc84
parent30308fd6e4f6f142ccfde96bc69348bca6574847
Rx: Treat rx_minPeerTimeout not as a minimum but as padding

An improved RTT and timeout calculation algorithm is being
developed but until we have it, treat rx_minPeerTimeout not as
a minimum value for the timeout but as padding to be added to
the measured RTT when computing the peer timeout value.

With this change rx does not begin to send large numbers of
resends when the RTT begins to exceed the rx_minPeerTimeout
value.  Timeout triggered resends at the moment can force rx
into fast recovery mode which in turn kills performance.  It
is better to avoid that problem for now.

Reviewed-on: http://gerrit.openafs.org/3026
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Tested-by: Jeffrey Altman <jaltman@openafs.org>
Reviewed-by: Jeffrey Altman <jaltman@openafs.org>
(cherry picked from commit 5c8981e0404287f2aa46390d3a9e8a52261906ab)
Change-Id: I82cf73cc8153d9aba669dc581231a88949b6ec4d
Reviewed-on: http://gerrit.openafs.org/3126
Tested-by: Derrick Brashear <shadow@dementia.org>
src/rx/rx.c