Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
/ 2005-06-13 15:34:59 +0400 \ Igor Yu. Zhbanov: > Hello! > > In some situations such as split-brain it is hard to predict syncronization > order of DRBD. if you know what happened, you can predict. if you don't know, you can use the "read_gc.pl" script (testing directory of the tgz) to read the generation counters on both nodes, and you know again... > So is it possible to ask DRBD to test whether it will be SyncTarget or > SyncSource? not directly, only by reading the generation counters by hand/script. btw, a running Primary will refuse to become SyncTarget, and rather go StandAlone. > Or is it possible to specify synchronization order > without invalidating the whole device? not directly, but by using the "--human" flag when appropriate, or by carefully manipulating the generation counters by hand. > It will help to recover after split-brains, so more recent data will not be > lost and total synchronization of 1.2 TB disk array will be avoided. well, after a split brain situation, you have to recover by hand anyways, and you will lose data, anyways. > If it not already possible, I think it will be two very useful features of > cluster management: > 1) Probing a synchronization order without starting the synchronization. no. > 2) Explicit chosing the synchronization order without total invalidation, > so sychronization can be quick. no. But. you recognize a split brain situation. - make sure that both nodes are disconnected, both# drbdadm disconnect all - stop heartbeat and/or services and monitoring thingies both# <whatever> stop - make the one with the worse data secondary worse# drbdadm secondary all (still disconnected!) - make the one with the better data secondary better# drbdadm secondary all (still disconnected!) - make the one with the better data primary, using the --human flag better# drbdadm -- --human primary all (still disconnected!) - reconnect the devices both# drbdadm reconnect all - start heartbeat and/or services ... both# <whatever> start should pretty much work as good as it gets. -- : Lars Ellenberg Tel +43-1-8178292-0 : : LINBIT Information Technologies GmbH Fax +43-1-8178292-82 : : Schoenbrunner Str. 244, A-1120 Vienna/Europe http://www.linbit.com : __ please use the "List-Reply" function of your email client.