[DRBD-user] Re: drbd-user digest, Vol 1 #82 - 6 msgs

Lars Ellenberg Lars.Ellenberg at linbit.com
Sun Mar 28 12:59:58 CEST 2004

Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.


/ 2004-03-28 09:54:47 +0200
\ Alessandro Marzini:
> 
> > Thanks Lars for your quickly response,
> > I'm going to do some test now.
> > BTW you have a drbd fan in italy! :D
> > I hope to help this software to growth in future.
> 
> Also me I'm an happy italian user of drbd!!
> 
> I use slackware 9.1 with kernel 2.4.24 and the other heartbeated machine
> with slackware-current, kernel 2.4.25. The two machine use the last 6.12
> drbd and the last drbd.
> I start drbd and heartbeat at startup with rc.local:
> 
> losetup /dev/loop0 /root/sito_netraid.bin
> losetup /dev/loop1 /root/database_netraid.bin
> /etc/rc.d/drbd start
> /etc/rc.d/heartbeat start
> 
> Actually I'm testing with loopback interface...
> and almost work well at 90%.

90% is missing the H of HA, I think.

For the non-working 10%, please report:
what exactly are you doing, step by step,
what did you expect,
when and how did real world deviate from your expectations,
and why do you think this is not how it should be.

DRBD on top of /dev/loop is not the best idea.
with recent kernels it may work.
On the other hand, it still may deadlock
under certain circumstances.

There is this parameter "diable_io_hints" which has the description
"Necessary if loopback devices are use for DRBD."
But this means the lo NIC, i.e. you have only one box, and let two DRBDs
on the same box talk to themselves. Together with sysrq P and T, this
gives nice stack dumps for certain strange corner cases. This is for
*me* only, for easily testing some weird setups without UML or real
hardware.

/dev/loop* as lower level devices I must strongly recommend against.

> Some time i noticed that one of two machine doesn't umount the nb0
> and nb1 mountpoint.

This sentence is not clear. I hope you do not mount one drbd on two
boxes at the same time. If you do, STOP THAT NOW. It will kill your
data, and sometimes your servers, too.

In any case, there should be something in the syslog,
maybe about "trying to uount ... cannot umount... busy..." .

The exact error is probably helpful.

	Lars Ellenberg



More information about the drbd-user mailing list