Unix Administrator writes:
> Hi,
>
> We got two AlphaServer 4100, each one with two KZPSA connected to an
> HSZ50 which resides on an SW800 cabinet :
>
> SW800:
>
> BUS C: Node A KZPSA -------HSZ50---------KZPSA Node B
> BUS D: KZPSA -------HSZ50---------KZPSA
>
> The first KZPSA-HSZ50-KZPSA is bus C, and the second
KZPSA-HSZ50-KZPSA
> is bus D.
>
> Actually, node A is accessing all the 22 disks on SW800 (through
buses C
> & D), and node B just access its own disks on SCSI local buses
(however,
> it can see all the 22 disks also). Node A has these disks configured
> with LSM on a diskgroup different than rootdg.
>
> What we plan to do (if possible) is to use B as a contingency node of
> node A. We've tried to deport/import the diskgroup, but it doesnt
works.
>
> In that sense, we plan to add the SW800's disks to the node B LSM
> database (with "voldiskadd rzb17", for example).
>
> Is this safe ? Will the voldiskadd operation initialize the disks ?
Any
> previous experience ?
>
> We know that this could be done with DECSafe ASE, but we would like
to
> know whether this works or not.
>
> Thanks in advance
_at__at__at__at__at__at__at__at__at__at__at__at__at__at__at__at__at__at__at__at__at__at__at__at__at__at__at__at__at__at__at__at__at__at__at_
%%%% The solution so far :
%%%%%%%%
_at__at__at__at__at__at__at__at__at__at__at__at__at__at__at__at__at__at__at__at__at__at__at__at__at__at__at__at__at__at__at__at__at__at__at_
Hi,
Many thanks to:
osf_at_python.corning.com
rjackson_at_portal.gmu.edu
sjirik_at_citipower.com.au
sxcak_at_java.sois.alaska.edu
chris_at_lagoon.meo.dec.com
eric_at_compgen.com
graham_at_sysnet-consulting.com
sysdfm_at_dtir.qld.gov.au
davis_at_decatl.alf.dec.com
Some conclusions:
1. This configuration is not officially supported by DEC.
2. This was not what we were trying to do, but several messages pointed
out that it is not possible to access simultaneously to one disk from
more than one node. DECsafe doesn't allow that. I'm not sure of what
happens with TruCluster, but as i understand, TC uses a sort of lock
manager, which would allow to access "simultaneously" to the same data
from different nodes, taking advantage of the very fast Memory Channel
interface.
What have we tested:
a. We added a brand new RZ29 to the SW800 cabinet, init'd and added to a
new LSM diskgroup "testdg" on node A.
b. Created a filedomain on the new volume ( mkfdmn /dev/vol/testdg/TEST
test_dom)
c. Copied some data to the fileset
d. umount'd the fileset and disabled the volume.
>From node B
a. cd /dev --> ./MAKEDEV rzg17
b. voldiskadm --> option 10 ( online a disk device ) or voldisk online
rzg17
c. option 8 ( access to ( import ) a diskgroup ) --> imported diskgroup
testdg
d. created by hand the /etc/fdmns/test_dom directory ( as in node A )
e. created by hand the link to /dev/vol/testdg/TEST
f. mounted the test_dom#test fileset on some directory on node B.
This worked fine. Now we can access the data that was created on node A
from node B. Bringing the data back to A was as simpleas reversing the
previous steps.
So far, this is the solution, but we plan to write some scripts in order
to automate the operation. We always keep in mind that we'll never try
to access the data simultaneously from nodes A & B.
Hope this helps.
Best regards
Received on Fri May 23 1997 - 01:48:14 NZST