Thread: Solaris 10 ZFS Postgresql request for comments

Solaris 10 ZFS Postgresql request for comments

From
Reid Thompson
Date:
Hi all,
I'm querying for feedback/comments.  Wondering what the list thinks of
the following.

Assume this is to provide a production database for a small company or a
department.  Production hours 5am-9pm for the most part so night-time
downtime if/when necessary would not be a problem.

Platform:
  SUN ultra 20 or intel/amd based PC
  Hard drive1      = OS Solaris 10(+)
  Hard drives2 & 3 = ZFS mirrored pool with PostgreSQL intalled
  Hard drives4-N   = ZFS raided PGDATA


Would this be considered viable?
Has anyone implemented anything similar?
Any obvious pitfalls that anyone is aware of?


Thanks,
reid

Re: Solaris 10 ZFS Postgresql request for comments

From
Christopher Browne
Date:
After a long battle with technology, reid.thompson@ateb.com (Reid Thompson), an earthling, wrote:
> Hi all,
> I'm querying for feedback/comments.  Wondering what the list thinks of
> the following.
>
> Assume this is to provide a production database for a small company or
> a department.  Production hours 5am-9pm for the most part so
> night-time downtime if/when necessary would not be a problem.
>
> Platform:
>   SUN ultra 20 or intel/amd based PC
>   Hard drive1      = OS Solaris 10(+)
>   Hard drives2 & 3 = ZFS mirrored pool with PostgreSQL intalled
>   Hard drives4-N   = ZFS raided PGDATA
>
>
> Would this be considered viable?
> Has anyone implemented anything similar?
> Any obvious pitfalls that anyone is aware of?

This is, in principle, a pretty generic sort of configuration that
ought to be fine for a not-overly-heavily loaded departmental sort of
application.

We have much smaller systems than that in production for applications
that don't take much load.

There are also much bigger systems to cope with cases that are
definitely bigger than what you describe.

You haven't described the nature of your application, so there is no
particular reason to believe this hardware to either be adequate or
inadequate...

Something worth considering is that if you use replication (e.g. -
something like Slony-I), you could quite likely live with having
several cheaper boxes.  Individually, they may not be as reliable as
the pricier one, but in aggregate, you should, via techniques like
"moving master," keep higher uptimes than a single cheap box could
support.
--
"cbbrowne","@","gmail.com"
http://cbbrowne.com/info/slony.html
"There  is no  psychiatrist in  the world  like a  puppy  licking your
face."  -- Ben Williams

Re: Solaris 10 ZFS Postgresql request for comments

From
"Jim C. Nasby"
Date:
On Thu, Mar 02, 2006 at 12:30:32PM -0500, Reid Thompson wrote:
> Hi all,
> I'm querying for feedback/comments.  Wondering what the list thinks of
> the following.
>
> Assume this is to provide a production database for a small company or a
> department.  Production hours 5am-9pm for the most part so night-time
> downtime if/when necessary would not be a problem.
>
> Platform:
>  SUN ultra 20 or intel/amd based PC
>  Hard drive1      = OS Solaris 10(+)
>  Hard drives2 & 3 = ZFS mirrored pool with PostgreSQL intalled
>  Hard drives4-N   = ZFS raided PGDATA
>
>
> Would this be considered viable?
> Has anyone implemented anything similar?
> Any obvious pitfalls that anyone is aware of?

I'd recommend drive 1 & 2 be mirrored witheverything but the table data,
which would go on a raid 10 of the rest of the drives via a tablespace.
That way you won't lose the box if drive 1 fails. You also probably
don't want raid5, if you were thinking about that...
--
Jim C. Nasby, Sr. Engineering Consultant      jnasby@pervasive.com
Pervasive Software      http://pervasive.com    work: 512-231-6117
vcard: http://jim.nasby.net/pervasive.vcf       cell: 512-569-9461