lars.ellenberg at linbit
Jul 25, 2012, 12:41 PM
Post #2 of 3
On Wed, Jul 25, 2012 at 04:11:42PM +0200, Arne Redlich wrote:
Re: Expected behaviour of "on-io-error pass_on"
[In reply to]
> There seems to be a disagreement between the documentation and the
> actual code wrt the "pass_on" I/O error handling strategy in drbd 8.4.
> Quoth http://www.drbd.org/users-guide/s-configure-io-error-behavior.html:
> "2. pass_on This causes DRBD to report the I/O error to the upper
> layers. On the primary node, it is reported to the mounted file
> system. On the secondary node, it is ignored (because the secondary
> has no upper layer to report to)."
> The code however suggests that reads are redirected to the secondary
> if the primary returned an error, and errors are not reported to the
> upper layers unless the secondary cannot satisfy the read request
> Bug in the code or in the documentation?
Bug in user's guide, copied and recopied from old man pages.
equivalent merge commit is in 8.4 as well.
The "passes error on to upper layers" was last describing the code
almost correctly for drbd 0.7, and for failed READs, only...
And that setting, even though "default" for much too long,
was not very useful back then already.
What is it you are trying to achive, that you even noticed this?
: 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.
drbd-user mailing list
drbd-user [at] lists