On Tuesday, February 12, 2019 10:15:10 PM MSK Konstantin Osipov wrote:
> * Georgy Kirichenko <georgy@tarantool.org> [19/02/12 17:09]:
> > Spare some vclock_copy invocations because they could be expensive.
> 
> OK, you did a full-scale diff & merge. This could actually be
> slower than the fast vclock_copy() which is available from the
> previous patch. Let's discuss.
I think in a typical case we wouldn't have more than 1-2-3 active masters so
it could be even faster especially when we have more than 3-4 cluster members.
Also it could be worth if I make some bench for that.