[DRBD-user] BAD! BarrierAck #422889 received, expected #422893

[Lars Ellenberg]

On Wed, Aug 22, 2012 at 02:57:06PM +0200, Lutz Vieweg wrote:
> Hi,
> 
> today we had to reboot a machine that had an involuntary power-outage,
> drbd version running (on both sides): 8.4.2rc1 (api:1/proto:86-101)
> GIT-hash: 1c425b5af957cead7753f974d7c4dae737fd2b14
> (the restarted machine had the role as "secondary" for 3 drbd devices).
> 
> The boot process stalled for about 5 minutes with drbd constantly
> emitting messages like this:
> 
> d-con ResourceData2: BAD! BarrierAck #422870 received, expected #422875!

Yes, I'm aware of that, and think I have fixed it finally.

> d-con ResourceData2: peer( Secondary -> Unknown ) conn( SyncSource -> ProtocolError )
> d-con ResourceData2: asender terminated
> d-con ResourceData2: Terminating asender thread
> d-con ResourceData2: Connection closed
> d-con ResourceData2: conn( ProtocolError -> Unconnected )
> d-con ResourceData2: receiver terminated
> d-con ResourceData2: Restarting receiver thread
> d-con ResourceData2: receiver (re)started
> d-con ResourceData2: conn( Unconnected -> WFConnection )
> d-con ResourceData2: Handshake successful: Agreed network protocol version 101
> d-con ResourceData2: conn( WFConnection -> WFReportParams )
> d-con ResourceData2: Starting asender thread (from drbd_r_Resource [2998])
> 
> After those minutes the synchronization seemed to work fine, and the
> system is now up and running.
> 
> I've read elsewhere that these messages might be "over-paranoid",

No, they are not.
Not in this case, at least.

> but if they hadn't stopped at some point, the boot procedure would
> have stalled like forever.
> 
> Can this be fixed?

This is fixed here.
Unfortunately I'm working on some more issues (one being in-kernel list
corruption) before tagging and pushing rc2.

Should happen later today/tomorrow, though.

Thanks for testing...

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