[Veritas-ha] Veritas VCS/DBAC 3.5MP1 & Kashya KBX4000
duncan.bradey at veritas.com
Wed Sep 8 21:16:04 CDT 2004
Looks like I replied too early :)
From: James T. Wonder [mailto:wonder at aip.org]
Sent: Thursday, 9 September 2004 3:21 AM
To: Martin White
Cc: Doug Hughes; veritas-ha at mailman.eng.auburn.edu; wonder at aip.org
Subject: Re: [Veritas-ha] Veritas VCS/DBAC 3.5MP1 & Kashya KBX4000
We are currently using Kashya with VCS, and are about to utilize it for
a clustered filesystem. We are told there is no issue with clustered
filesystems and that Kashya have tested with them. They mentioned that even
with multiple hosts writing to the same LUN they maintain write order
consistency. After working with them I believe anything they promise, I
have had no issues with their product.
The units work *VERY WELL* and I cannot say enough about the ability of
the company to work with us. We tried other products (will not name
names) prior to using the KBX4000, and these units are by far the easiest to
The KBX system sits on the SAN and is mapped to the LUNs on the storage
units, one set near end and one set far end. The Kashya boxes are clustered
on each side and maintain communication over your WAN connection.
The splitter gets installed on your hosts (we run Solaris 8) and it runs
a daemon as well as loads a module. Where in the stack it sits I did not
ask, but it basically sends the written block to both the disk and the
Kashya box. The KBX4000 gets mounted as a disk on your system and the
splitter writes to the LUN that is presented.
If you need anything else, please do not hesitate to ask.
Doug Hughes wrote:
> On Wed, 8 Sep 2004, Martin White wrote:
>>We are currently running VCS/DBAC 3.5MP1 for Oracle 9i RAC under
>>Solaris8, there are two nodes in the cluster. The storage for the
>>cluster is provided by an IBM ESS connected via McData directors, the
>>hosts using Emulex LP9002 cards to connect to the SAN.
>>One database on the cluster runs in Active/Passive mode and so makes
>>use of a clustered file system. All the other databases run in
>>traditional VCS failover mode and do not use a clustered filesystem.
>>At some point all databases will be migrated to active/active,
>>probably when we re-build the cluster next year on the latest releases of
all the software.
>>Currently Disaster Recovery is handled by Oracle replicating its data
>>to a non clustered server at the DR site.
>>On other servers we use Veritas Volume Replicator to keep DR in step,
>>other servers just use simple rsync setups. The Wintel team do
>>something else with their servers.
>>Now the company I work for want to implement a single replication
>>method and have their eye on the KBX4000 from Kashya
>>(http://www.kashya.com/). Now the way this works, so I'm told, is that
>>you install a driver on all hosts that you want to replicate, which
>>sits somewhere in the I/O stack (not sure where
>>yet) and splits off a copy of the data which it then passes to the
>>Now I can see that wouldn't be a problem with a single server. But
>>once VCS is introduced I can see things getting a bit complicated, and
>>as for VCS/DBAC I'm not sure how kashya will handle writes to the same
>>LUN from more than one host at the same time.
>>Anyway, has anyone here had any experience of the KBX4000 with VCS or
>>VCS./DBAC, if so what do you think of the product, are there any
>>pitfalls or does it just work? If not, what do you use for replication
>>other than Volume Replicator that works well in a cluster environment.
> This isn't an answer, but, what are the advantages the KBX4000 has
> over VVR? cost? Anything else?
> A lot of people have been talking about EMC's replication technology.
> Apparently it's becoming slightly less vendor specific.
> Veritas-ha maillist - Veritas-ha at mailman.eng.auburn.edu
Veritas-ha maillist - Veritas-ha at mailman.eng.auburn.edu
More information about the Veritas-ha