Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
/ 2006-07-04 22:21:05 +0800 \ Badge Parag-G19470: > From your last point: > "But in case there are blocks with data divergence that are not covered > by the bitmap, all would apear to work, until the next failover and > maybe even beyond, and then suddenly out of nowhere BOOM you notice data > corruption. and you'd blame it on drbd. which would be kind of unfair." > > I think you are suggesting that it is better to do full sync to avoid > possible data corruption on fail-over? If you have any doubt about which blocks may be different on the two nodes, you have to do a full sync. That just the facts. Otherwise you'd have to live with the possibility that DRBD thinks it is bitwise identical on both nodes, since you lied to it, but in fact there are a few blocks that _are_ different, which drbd does not know about. These differences would only show up on the next failover, since reads are done local where possible. And depending on where these blocks are this may be completely harmless (some differing data block in the unallocated fs space) or catastrophical (important block in some data base journal). So: if you're in doubt, do a full sync. -- : 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.