Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
On Tue, Sep 18, 2007 at 12:56:40PM -0700, Kelly Byrd wrote:
>
> I'm testing out a single drbd in pri/pri with a single volume group
> containing many logical volumes. The LVM scheme is static, I set it up
> once, it never changes when the drbd is live. I then ext3 format each LV. I
> mount half the LVs on one box, half on the other. On failure, I STONITH and
> mount the "other" LVs on the remaining box. Does this seem reasonable? LVM
> isn't changing underneath anyone and ext3 is only ever accessed by a single
> host at a time.
if it is static, anyways,
why not have drbd on top of lvs,
and run "classical" Primary/Secondary per drbd/lv?
pri/pri does basically work,
but is more complicated to handle.
if you go for pri/pri in your case,
for the setup described here,
after thorough contemplation,
you may want to configure the "after-split-brain handlers" to
something similar:
if secondary connects to primary (1pri)
and the detect divergent data sets:
discard the changes on secondary, sync primary -> secondary
if primary connects to primary (2pri)
disconnect. aparently stonith did not work.
sort out manually.
if secondary connects to secondary (0pri)
disconnect.
aparently this was a cluster reboot/common power cycle.
sort out manually.
--
: Lars Ellenberg http://www.linbit.com :
: DRBD/HA support and consulting sales at linbit.com :
: LINBIT Information Technologies GmbH Tel +43-1-8178292-0 :
: Vivenotgasse 48, A-1120 Vienna/Europe Fax +43-1-8178292-82 :
__
please use the "List-Reply" function of your email client.