Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
On Thu, Aug 09, 2007 at 07:40:04AM -0500, James Wilson wrote: > Thanks for the reply. I haven't had this happen on its own. I was just > testing by pulling the network cable to see if everything would failover > and noticed it when the other server came back up. Also to manually > resync them I just issue the command drbdadm invalidate "r0" or whatever > your resource is. Is this the wrong way to do it? well, by "invalidating" you force a _full_ sync. the "discard-my-data" thing still does an incremental sync, omitting those blocks of which we know for sure that they did not change. [...] > >you could configure some "auto-recovery strategies", see after-sb-#pri > >settings in drbd.conf. > > > >to get them reconnected and resynchronized manually, > >you'd have to typically > > chose one side to do > > umount > > drbdadm -- --discard-my-data connect XY > > and maybe on the other side do normal > > drbdadm connect XY > > > >we are going to make this less annoying than it is now by introducing a > >write_quorum and IO-freeze on disconnect, then we can timeout for a > >possible imminent reconnect (network hiccup only), and after that > >timeout arbitrate (triggered by the cluster manager) which side may > >continue, and which has to be fenced (or let it self-fence...). > > > >but for the time being, yes, I know, two-primaries is still "inconvenient" > >when you have a flaky network. -- : Lars Ellenberg Tel +43-1-8178292-0 : : LINBIT Information Technologies GmbH Fax +43-1-8178292-82 : : Vivenotgasse 48, A-1120 Vienna/Europe http://www.linbit.com : __ please use the "List-Reply" function of your email client.