[DRBD-user] documentation clarification - Generation Identifiers

Lars Ellenberg lars.ellenberg at linbit.com
Thu Jun 6 14:44:13 CEST 2013

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, May 30, 2013 at 05:52:45PM -0400, Jonathan Woodbury wrote:
> Hi there,
> 
> In reading this page (http://www.drbd.org/users-guide-8.3/s-gi.html) your
> online docs, in the "How DRBD uses generation identifiers" -> #5 "Current
> UUID matches peer's historical UUID" section, I found something a bit
> confusing and I'm suspicious that it may be incorrect.
> 
> It states:
> 
> "The local node detects that its current UUID matches one of the peer's
> historical UUID's. This implies that while the two data sets share a common
> ancestor, and the local node has the up-to-date data, the information kept
> in the local node's bitmap is outdated and not useable. Thus, a normal
> synchronization would be insufficient. DRBD now marks the entire device as
> out-of-sync and initiates a full background re-synchronization, with the
> local node becoming the synchronization source."
> 
> If the local node's current UUID matches one of the peer's historical UUIDs
> wouldn't that mean that the local node's data is an older generation than
> the peer's and the local node's data should be considered out-of-date and
> replaced with the peer node's data?

Absolutely.
Documentation bug.
In that paragraph(s), source and target need to be swapped.

DRBD checks that
 peers are not identical,
 peers used to be identical at some earlier point in time,
 one peer stopped changing anything while the other evolved,
 which implies that ther is no data divergence
  (which would usually be the fall-out of some "split-brain"),
 and that unfortunately something happened that made a bitmap based
 resync unreliable/impossible.

This should not happen, but sometimes it still does.
(or can be provoked by a specific sequence of events)

DRBD does "the right thing" (use the source to verify...).
(( speaking of source, there are of course even more corner cases
   handled than are documented in the "prose" ))

> Also, thanks for providing and hosting these high quality docs!  I hope
> that this discussion can serve to help make them even better or set me
> straight ;)
> 
> -- 
> Jonathan Woodbury
> Ganeti SRE - NYC
> Google Inc.

-- 
: Lars Ellenberg
: LINBIT | Your Way to High Availability
: DRBD/HA support and consulting http://www.linbit.com

DRBD® and LINBIT® are registered trademarks of LINBIT, Austria.



More information about the drbd-user mailing list