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, Jul 15, 2010 at 09:58:13AM +0200, Lars Ellenberg wrote: > On Thu, Jul 15, 2010 at 09:16:43AM +1000, adamn at eyemedia.com.au wrote: > > Ok I didn't realise this was that kind of an issue, I just assumed DRBD > > wasn't happy with my setup due to the second node not being online. > > The partition /dev/sdb1 is only set to about 40GB while I'm testing the > > setup, once I've nutted everything the machine will be rebuilt to have > > it's full capacity. (around 7TB) it's an LVM partition > > > > I created a single Logical Volume around 20GB have it exported via > > iscsi-scst. > > > > Single VMWare ESXi server (again this is all testing this phase), setup > > iscsi to connect to the SAN (testing MPIO) as soon as I "scan" from the > > storage adapter I start getting the > > [28609.797320] block drbd1: al_complete_io() called on inactive extent 623 > > > > ete.. > > > > see http://www.genis-x.com/files/syslog > > I never encountered that particular message, > not in testing, not in production. > I cannot remember any report of it, either. > > It indicates serious imbalance in reference counts of "active extents". > > So "something" is "special" with your setup. > I suggest some sort of build inconsistency, > so please double check a clean build of the > DRBD module exactly fitting the kernel. > > Try to narrow it down. Try to reproduce without going through the iSCSI > layer first, accessing DRBD locally. Try to reproduce with a "vanilla" > kernel, using a different iSCSI target. Or, of course, contact sales at LINBIT, checkout a suitable package, and have us investigate and fix it. ;-) Of course, DRBD issues get eventually fixed anyways, once we can reproduce them. Currently we cannot, though. -- : 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. __ please don't Cc me, but send to list -- I'm subscribed