Date: Mon, 30 Jul 2001 09:47:23 -0400
Thanks to everyone who replied. Here are the responses I received: -
* This message was sent from Geocrawler.com by "Brandon Addison"
<addison_at_alf.cpqcorp.net>
Data Replication Manager is the supported utility
for keeping two separate/identical StorageWorks
cabinets constantly synced. This entails a very
complex setup, so calling the storage team at
Compaq is the best bet.
You can replicate over fiber or ATM, so distance
is not an issue.
see
http://www.compaq.com/products/storageworks/Storag
e-Management-Software/DataRepProd.html for more
info.
* You should be able to synchronize the data between the HSG's by
configuring the LSM (Logical Storage Manager) software add-on to perform
host based mirroring on both cluster members (assuming that's the role of
the ES40's).
See the attached link for reference.
http://tru64unix.compaq.com/faqs/publications/base_doc/DOCUMENTATION/V51_HTM
L/ARH9BBTE/TITLE.HTM
Regards,
Paul (McDowell, Paul [Paul.McDowell_at_celera.com])
* What do you mean synchronozation of data.Do u want to have same data
both MA8000 ? If so then you to set up script or use rdist to copy the data
..rdist will copy the data and even compare the old files if it exits and
copies only files which are changed
Regards
(Manish Vashi [manish_vashi_at_fanniemae.com])
* FYI:
There are many Compaq supported solutions, at least half a dozen.
They vary in how fast you need it to be (every block, every few seconds,
every hour)
How far apart the HSGs are (i.e. feet, miles, hundreds/thousands of miles).
How synchronized you need the data. (i.e. block by block, record by record,
transaction by transaction, file by file).
For really slow, across the world, and file by file, rdist will work, but
you probably know that.
If it is all Database records and fields, you may want to just use the
Databases Replication feature.
With an optional package, the HSGs internally support both synchronous and
asynchronous block by block replication as well as other replications.
For kernel based immediate replication where you don't want to complete the
write() call until it is written and make sure fsync() type stuff works, I
would look into LSM (Logical Storage Manager) and it's mirroring capability.
My guess is that you want one of the HSG internal packages, but even then I
think there are several choices.
BTW: If you still have questions in 2 weeks, feel free to e-mail me
directly. I will have been to a 1 week storageworks class by then and the
internal packages should be covered pretty well during that class.
Greg Freemyer
Internet Engineer
Deployment and Integration Specialist
The Norcross Group
www.NorcrossGroup.com <http://www.NorcrossGroup.com>
* Do you mean .... your trying to enable a Data Replicator Manager
Environment?
if so.... take a look a this :
I think you need a connection (FC) between the two HSG80 controllers
and special software to replicate data (by hardware).
(Raul Sossa S. [RSossa_at_datadec.co.cr])
* Is just one system responsible for writing? If more than one, is
it
the two systems on the fabric or any of the systems on the network.
Using the clustering support in V5.1 you could make the Database
and Appliation servers two members of a cluster. Then they'd be
able to completely share the attached storage. "Syncronizing"
data across the two HSG80 would be as simple as using LSM to
mirror the data across the two subsystems. Though, by making
the two a cluster the question may not be as relevant.
V5.1 clustering is limited to systems that can be connected by
Memory Channel. I don't know if DS20s support Memory Channel.
V5.1A clustering is rumored to allow using the LAN for cluster
communciations.
If just one system is writing, then using the failover services
of the previous version of clustering might be sufficient. Make
the systems NFS servers of the data, with the writer the primary
source of the service. If it fails, the service fails over
and the other can provide access. Again "synchronization" can
be with LSM to just mirror the data.
If the data on the HSG is under a database and on raw devices
instead of a file system, the particular database system may
allow shared access when the storage is on a shared interconnect
(Fibre Channel in this case).
My previous message mentioned DRM which is the Data Relication
Manager. It is designed to provide disaster tolerance over
moderately long distances (10s of km). I think the organization
is that one port of each HSG is connected to an ATM network
and the two controller pairs can be separated by a considerable
distance. One controller is the master and pushes all writes
over the ATM connected ports to the other. A system at the
remote site can be a hot standby in case the master fails.
You don't seem to need the distance.
( Alan.Rollow_at_compaq.com <mailto:Alan.Rollow_at_compaq.com> )
* Consider this suggestion...
Don't view the HSG80 as the component that has to be
isolated to a specific system. View the individual
logical units as the thing that can be isolated.
Doing this you can have logical units (disks as the
host sees them) that are spread among the HSG (and
any suitable Fibre Channel connected storage). If
you want failure protection that can survive the loss
of a given storage subsystem mirror the data among
units on as many subsystem as you can afford. In your
case, two-way mirroring.
LSM on the hosts can do the mirroring and it doesn't
care where the unit is as long as it behave like a
local disk (which is what the HSGs present).
For failover on V4 you can use the TruCluster software.
On V5 (starting at V5.0A) you can use that version of
TruCluster to completely share any file system that
are available to either host.
( Alan.Rollow_at_compaq.com <mailto:Alan.Rollow_at_compaq.com> )
Regards
Raymond Lewis
ILLUMINAT
Phone: 1.868.622.4551
e-mail: raymond.lewis_at_ILLUMINATnm.com
(image/jpeg attachment: image001.jpg)
(image/gif attachment: image002.gif)