what network-FS for web-cluster?

Hi,

Whats the best network-FS for a Linux LVS cluster used as a web-server
for a (mysql)DB application?
I can use no network-FS and replicate the DB but then I have a problem
with session-management and also if the DB gets really large I run
into problems with that too. I think using persistent connections is
ugly so I would like to use a network-FS to at least store session-
data, but would better like a network-FS to store the entire DB.
The main DB will be large and holding lots of photos, video, and text.
Allmost all web-page requests will require DB reads, only a few will
require writes (except for the session-management data).
I want very high-availability by the way, 100%.

I looked into Lustre and PVFS, but they don't seem ideal, they get
ugly when you need failover. Not sure about NFS.

Is there no network FS that can just sort of create a RAID? I will use
comodity servers for the cluster. Each has a single disk, I want to
stripe and mirror over the disks to get both redundancy as well as
extremely high capacity. Whats the solution for this? If a server goes
down, everything must continue as normal without any hickups!

Thanks alot!
Tobias

On Jul 25, 8:22 pm, nsa....@gmail.com wrote:
> Hi,
>
> Whats the best network-FS for a Linux LVS cluster used as a web-server
> for a (mysql)DB application?
> I can use no network-FS and replicate the DB but then I have a problem
> with session-management and also if the DB gets really large I run
> into problems with that too. I think using persistent connections is
> ugly so I would like to use a network-FS to at least store session-
> data, but would better like a network-FS to store the entire DB.
> The main DB will be large and holding lots of photos, video, and text.
> Allmost all web-page requests will require DB reads, only a few will
> require writes (except for the session-management data).
> I want very high-availability by the way, 100%.
>
> I looked into Lustre and PVFS, but they don't seem ideal, they get
> ugly when you need failover. Not sure about NFS.
>
> Is there no network FS that can just sort of create a RAID? I will use
> comodity servers for the cluster. Each has a single disk, I want to
> stripe and mirror over the disks to get both redundancy as well as
> extremely high capacity. Whats the solution for this? If a server goes
> down, everything must continue as normal without any hickups!
>
> Thanks alot!
> Tobias

this isn't a database question, but a clustering & file system
question.

You might get better responses in a comp.os.linux.* group.

Good luck!
Ed