From: NetComrade (netcomrade@bookexchange.net)
Date: Fri Jun 10 2005 - 14:33:06 EDT
So we've been waiting for Veritas to start supporting AMD chips in 64bit
mode, so that we could migrate off Sparc II to Opteron machines. Veritas
finally announces support:
Apparently, Veritas for Solaris x86 (x64) will be crippled. The cluster
server is not even planned yet (or so I was told), only Storage Foundation
will be available (which they tested with v40z and 3510fc)
The Linux version, on the other hand, will come in 'full force'. Veritas
Foundation will come with HA and even Oracle RAC support (we run
Oracle). However, they seem to be mostly supporting high-end storage
devices. When I told them that it seems to be counter-intuitive for folks
moving to cheaper Linux hardware, to use high-end storage, they argued
that "if you save so much money on buying servers, you should be able to
upgrade your storage to high end". What's 'low-end' with the 3510FC
(performance wise) they weren't able to explain.
You can see the Veritas Supported Hardware Compatibility List here:
http://ftp.support.veritas.com/pub/support/products/VolumeManager_UNIX/271281.pdf
Question: What if I do run an 'unsupported' disk array? What does Veritas
care? Will the DMP simply 'not work'? Shouldn't it it work with any
hardware, as long as Veritas determines that it's the same 'disk' from the
private region? Isn't that the whole point of the private region?
Our main reasons for sticking with Veritas have been DMP, Cluster Server,
some additional flexibility from Volume Manager vs Hardware Raid. The
biggest advantage of sticking with Veritas is of course that we know it.
So it seems, my only option, if I want to run Oracle on Opterons is to go
to RedHat Linux, and try to use the support of DMP in kernel 2.6 (the tool
is called mdadm), together with Red Hat Cluster, which I believe supports
Oracle database.
However, mdadm is error prone. One has to manually identify paths to
disks, and manually map them (or at least that's what it seems like). It
also looks like, if for whatever reason device names change (e.g. you get
a new controller), your multipathing will be broken, in fact, you might
actually corrupt data that way (I have no experience with mdadm, just
speculating here). Additionally, mdadm works in Active/Passive mode (there
is no IO load balancing).
You can read more about md support with Sun arrays here:
http://www.sun.com/storage/san/linux_multipathing.pdf
The doc mentions something about having 'MPIO' (multipathing
IO?) capability built in, but I am not really familiar with the
technology. Do these drivers allow to failover between ports within an HBA
or between HBA's as well?
Anyway.. had to vent a bit, and maybe hear some suggestions on the next
steps.
Thanks!
_______________________________________________
sunmanagers mailing list
sunmanagers@sunmanagers.org
http://www.sunmanagers.org/mailman/listinfo/sunmanagers
This archive was generated by hypermail 2.1.7 : Wed Apr 09 2008 - 23:30:52 EDT