[Drbd-dev] Understanding barriers in DRBD

[Tejaswini Poluri]

Hii all,

I have been trying to solve an issue on DRBD-8.3 branch since a month and
could not yet find a solution.  We have a iscsid secondary disk and a
primary drbd mounted disk using protocol B and are in sync.  During writing
in the primary when the network of the iscsi disk of secondary is down, the
request timer function is supposed to be triggered within ko-count time and
the primary should change the connection state to Timeout and then to
Network Failure state. This is happening in Protocol C and not is not
working in Protocol B.

During debugging, we came to stage where  even after the iscsi disk is in
blocked state the Primary sends *w_send_barrier* continuously even though
there is no *BarrierAck.*

I would like to understand the implementation of barriers and barrier acks
of DRBD in more detail. I think there could be some synchronization problem
of barriers for every epoch. It would be great if someone could share me
some pointers about barrier acks,


Regards,
Tejaswini
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.linbit.com/pipermail/drbd-dev/attachments/20151027/6e5351a0/attachment.htm>