Hi,
Just some comments on your situation , please take a look the following
things :
* Sometimes the hw looks the same, i'm talking specifically to the
SSD's, but they can be somehow different and that may lead to some
problems in the future . Bottom line is , don't go too cheap for
production :)
* If you bought 3 7210 full of disks, they put pressure in your Sun
reseller to get some very cheap (i would say for free!) logzillas
. Their margin is more than okay to give you at least 1 logzilla
to your primary box .
* With ZFS always make sure you use a UPS attached to system and
make sure that you make a gracefull shutdown of the system in the
event of power failure and drained batteries on the UPS. If
possible use some sort of no-break system, like an auxiliary
generator or so.
* If possible don't use any box whatsoever in production before 2 to
4 weeks of testing/pre-production . It's quite often to see
problems when we stop to use benchmarks and start to use the
systems in production. Benchmarks i often called them "controlled
chaos", but based on my experiences in production there's only
"chaos", without the "controlled" part ;) . We never now when some
application that *should* do some sequencial reads is actually
doing random reads...
* Don't give up "easily" on the Sun support guys . Again based on
my experiences, quite often you need to put pressure in your Sun
partner/reseller to get the necessary amount of attention to your
problem..please remember that money is what drives business .
* On your XEN, do you really LVM-over-iscsi? Can you live with NFS
storage repositorys in the XEN? In my current job, i never
achieved better performance with iscsi than what i have with NFS,
and NFS gives less issues..just take a look to the all snapshot
issues in Citrix Xen forums for people using ISCSI rather than NFS
. Also in a near-by future (i hope) pNFS will be there so,
therefore more and more performance for the NFS stack ;)
* Still on the XEN, i found that in my current environment enabling
jumbo frames in the xen physical servers leads to massive
problems..so if you can take a look to the jumbo frames support
for your specific network cards
* Getting out from the OS installed in your 7210, towards anything
else probably means that you will lose support from Sun..so i
don't know if that's such a good idea. Again, put pressure in the
Sun reseller to get attention to your problem and don't take
shortcuts and don't get mad ;) I been there, i know the feeling of
having to put something , expensive, to work and it just doesn't
work as it should.
* Keep this list informed, they are great minds around here, so
maybe someone can give you some extra help
Good luck,
Bruno
On 23-2-2010 23:45, Nate Carlson wrote:
> Hey all,
>
> I've put up a blog post on issues we're having with 3 brand new Unified
> Storage 7210 arrays, available here:
>
> http://tinyurl.com/yeo9pft
>
> I'm curious if the symptoms (spurious halt, spurious vlan ipmp group
> dropping, spurious reboots) that we are seeing are common for the X4540
> machine, or if you think it's likely to be a problem with the way Sun rolls
> the Unified Storage software package.
>
> If it is a software issue, and we are unable to return these machines, I'm
> considering running either OpenSolaris or NexentaStor on the hardware..
> hoping that it's a software issue and not hardware. ;) (Going OpenSol or
> Nexenta would also let us add our own SSD's without worrying about Sun
> complaining about it.. $400 for 32gb x25-e's sounds a lot better than $6k for
> an 18gb Logzilla!)
>
> Appreciate any thoughts!
>
smime.p7s
Description: S/MIME Cryptographic Signature
_______________________________________________ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
