On Thu, 17 Jan 2019 at 02:31, Igor Podlesny <[email protected]> wrote: > > On Thu, 17 Jan 2019 at 02:25, Ondrej Zajicek <[email protected]> wrote: > > On Thu, Jan 17, 2019 at 12:38:11AM +0700, Igor Podlesny wrote: > > > Or would the second "config timeout" commit version 2 instead and when > > > timeout is reached would rollback from version 3 (which is equal by > > > content to version 2) to version 2 instead of version 1? > > > > Yes, this way. There is just one 'old configuration' slot that is used > > for rollbacks. > > Thanks! But birdc won't show any confirmation/warning that it's applied, > right?
To improve birdc giving its users more convenient and often _expected_ behaviour -- it's only needed to check on "config timeout" request if another postponed rollback had been already scheduled and if there were no config changes interim, just to postpone current rollback accordingly -- using freshly entered timeout value of course. Easy, clean and very convenient fix allowing to start with short interval of rollback and consequentially increase it -- kinda "manual anti-slow-start": allowing to rollback fast if changes led to lost of control almost immediately but also allowing to do auto rollback if lost of control happened a bit later. -- End of message. Next message?
