I discovered the problem exactly because I wanted to fail quickly, having a global fr_timer of 5 s.
IMHO the patch as it is doesn't affect the capability to fail quickly since a node is taken off the nodelist for message broadcasting immediately, except for the ping. I would say that it affects the recovery speed, since a two pings cycle is needed before a node is added back in case of a temporary failure.
A modparam can be added to enable this, along with the caveat in the doc that a race condition can happen due to the combination of fr_timer and dmq ping interval.
—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub, or mute the thread.