RE: [U2] Putting Universe on a SAN

2007-11-10 Thread David Wolverton
We have customers running UniData on a SAN and very pleased with the
results.  The snapshotting really isn't as practical as they'd have you
believe was probably the biggest negative surprise - their users were
unwilling to deal with a 15-30 second 'DBPAUSE' more than once or twice a
day, so the idea of doing a 'snapshot an hour' went out the window in the
first week.  Otherwise, the users don't notice the difference on the
LeftHand SANs using gigabit ethernet we've seen out there. Just make sure
you have the ability to prioritize and localize to 'hide' us from the
constant Exchange disk assault -- but most anything will do that now in
SAN-Land.  Additionally, we have one client totally virtual -- Vmware
Infrastructure and a SAN -- most of their end users are hard pressed to tell
the difference.

YMMV ... But if it's rationally assembled, and ultimately 'looks' like a
"Local" "D" drive to the UniData/UniVerse system - it can have good results
and allow you to 'hardware mirror' outside your four walls.  But shop
carefully  -- you'll want to go spindle happy. More is frequently better in
these setups as it allows 1+0 mirroring in the SAN as well.

David W.
---
u2-users mailing list
u2-users@listserver.u2ug.org
To unsubscribe please visit http://listserver.u2ug.org/


RE: [U2] Putting Universe on a SAN

2007-11-09 Thread Dan Fitzgerald
Well, this would be a not-so-good way to go.

The big sticking point in this particular scenario is that you're sharing
the SAN. I know, that's part of the point of having one. But what happens is
that UniVerse is requesting relatively small I/O from the OS, typically 4,
8, or 16K per request, and usually lots of those. Other things don't
necessarily do the same thing; a foxpro page could also bring up pictures
and very pretty graphics, so you may get a 1 Mb request from the SAN for
that. While it's servicing that 1Mb request, UV's 4K requests often are
piling up in the queue.

Then we have to consider thinks like how you've carved out your metas &
luns, stripe sizes (I've seen places with 960K stripe sizes, to service 16K
requests...), how many actual spindles have you got uv on, etc.

Steve O'Neil has been point on these issues, and is probably the fellow with
whom you want to speak.

-Original Message-
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] On Behalf Of Jerry Banker
Sent: Friday, November 09, 2007 2:02 PM
To: u2-users@listserver.u2ug.org; [EMAIL PROTECTED]
Subject: [U2] Putting Universe on a SAN

I know I saw some information on this list about not putting Universe on
a SAN. Does anyone have that information readily available? Does anyone
know if the problems they were having were fixed or could be fixed? We
are thinking of putting our Universe data accounts, not the runtime, on
a SAN and using replication to another SAN off site. There will be other
databases on the SAN, FoxPro and SQL Server, Will this cause a problem?
Are any of you doing this? What kind of results are you getting? UV is
on Red Hat the others are on Windows will that be a problem?



Jerry
---
u2-users mailing list
u2-users@listserver.u2ug.org
To unsubscribe please visit http://listserver.u2ug.org/
---
u2-users mailing list
u2-users@listserver.u2ug.org
To unsubscribe please visit http://listserver.u2ug.org/


RE: [U2] Putting Universe on a SAN

2007-11-09 Thread Jerry Banker
They're talking about mounting the SAN through a fiber channel instead
of NFS or Samba. Will the same problem apply.

-Original Message-
From: Richard Nuckolls [mailto:[EMAIL PROTECTED] 
Sent: Friday, November 09, 2007 3:13 PM
To: u2-users@listserver.u2ug.org
Subject: Re: [U2] Putting Universe on a SAN

re SAN's and Universe:

DELETE.FILE and RESIZE.FILE will not work on dynamic files that are  
NFS mounted. (Universe keeps the DATA.30 file open while trying to  
delete the enclosing directory.  This will fail on an nfs mounted  
directory.) Override the VOC entry with your own version.

Do not access a snapshot of a file from the same mount as the  
original.  At least on NetApps, the inode of the snapshot will  be  
the same as the inode of the original.  Universe will confuse the data.

Besides the usual caveats about making sure that a single machine  
controls locks for any given file, it is also worthwhile to keep the  
NFS under a set folders(4-6 or so) and then give each of these  
folders its own mount point on the host machine. Effectively creating  
multiple devices under UV.

Of course: do not try this without a good UPS.  Most SAN's have one  
built in too.

Corruption of a group is still possible, but our experience has been  
good.

My advice would be to replicate snapshots rather than live data  
(Unless you are using Uv Replication).  With snapshots, you can  
temporarily suspend IO, take a snapshot, and continue.

If you are doing live replication at the SAN level, then a major  
problem (an errant DELETE.FILE, for instance) will be replicated on  
your backup SAN.  Also, suspending IO will avoid catching files in  
transitional states.

Good Luck!

-Rick Nuckolls
Lynden, Inc

On Nov 9, 2007, at 11:01 AM, Jerry Banker wrote:

>  Does anyone know if the problems they were having were fixed or  
> could be fixed?

Not at 10.2.2 for those mentioned above.  I have not used Uv Data  
Replication.
---
u2-users mailing list
u2-users@listserver.u2ug.org
To unsubscribe please visit http://listserver.u2ug.org/
---
u2-users mailing list
u2-users@listserver.u2ug.org
To unsubscribe please visit http://listserver.u2ug.org/


Re: [U2] Putting Universe on a SAN

2007-11-09 Thread Richard Nuckolls

re SAN's and Universe:

DELETE.FILE and RESIZE.FILE will not work on dynamic files that are  
NFS mounted. (Universe keeps the DATA.30 file open while trying to  
delete the enclosing directory.  This will fail on an nfs mounted  
directory.) Override the VOC entry with your own version.


Do not access a snapshot of a file from the same mount as the  
original.  At least on NetApps, the inode of the snapshot will  be  
the same as the inode of the original.  Universe will confuse the data.


Besides the usual caveats about making sure that a single machine  
controls locks for any given file, it is also worthwhile to keep the  
NFS under a set folders(4-6 or so) and then give each of these  
folders its own mount point on the host machine. Effectively creating  
multiple devices under UV.


Of course: do not try this without a good UPS.  Most SAN's have one  
built in too.


Corruption of a group is still possible, but our experience has been  
good.


My advice would be to replicate snapshots rather than live data  
(Unless you are using Uv Replication).  With snapshots, you can  
temporarily suspend IO, take a snapshot, and continue.


If you are doing live replication at the SAN level, then a major  
problem (an errant DELETE.FILE, for instance) will be replicated on  
your backup SAN.  Also, suspending IO will avoid catching files in  
transitional states.


Good Luck!

-Rick Nuckolls
Lynden, Inc

On Nov 9, 2007, at 11:01 AM, Jerry Banker wrote:

 Does anyone know if the problems they were having were fixed or  
could be fixed?


Not at 10.2.2 for those mentioned above.  I have not used Uv Data  
Replication.

---
u2-users mailing list
u2-users@listserver.u2ug.org
To unsubscribe please visit http://listserver.u2ug.org/