Re: N+1 Configuration
- From: "Jasper Smith" <jasper_smith9@xxxxxxxxxxx>
- Date: Thu, 31 Aug 2006 19:17:45 +0100
Every clustered instance will require it's own drive letter but off that you
can hang a bunch of mount points. We tend to have
Mounts\SQLData
Mounts\SQLLogs
Mounts\SQLDumps
Mounts\SQLTemp
We have tested upto 16 instances on a 4 node cluster all using a single
drive letter and then mount points as above for each and not had any
problems
How to configure volume mount points on a Microsoft Cluster Server
http://support.microsoft.com/kb/280297/
--
HTH,
Jasper Smith (SQL Server MVP)
http://www.sqldbatips.com
"DTully" <DTully@xxxxxxxxxxxxxxxxxxxxxxxxx> wrote in message
news:997AEB50-0F2D-46D2-9DDC-718B60B0FE67@xxxxxxxxxxxxxxxx
Wow! Thank you both!
The team brought up the idea of using NTFS junction points (the same
thing?). Is there any good documentation of this? So it does appear that
no
two instances of SQL can be using the same drive letter though (even if,
as
Linchi suggested, mounting all other drives under that drive letter). We
have
consistency between servers that all data is on drive F, etc. But that's a
trade off for the ability to have fewer failover servers per grouping of
active servers.
Thanks!
Darin
"Linchi Shea" wrote:
To add to what Geoff said about mount points.
If you use one drive letter per SQL instance and mount all other drives
for
that instance under that drive letter, you can technically have 20+
instances
in the cluster. But hopefully, you are not considering that many
instances in
a single cluster.
I have tested various scenarios of using mounted volumes in a SQL2005
cluster. There is no major issue with using mounted volumes in SQL 2005.
But
note that some of the disk space related utilities are not 'mount point
aware', thus won't give you the total space info for the entire drive
tree or
may not be aware that there is a separate drive mounted underneath a
drive
letter. So you may have to revise your space monitoring setup.
Linchi
"DTully" wrote:
Thanks Linchi!
I think I see the light now. So one clustered OS instance with multiple
nodes, with multiple SQL installations, etc, etc.
Only problem I see then is each SQL named instance would have to use
different drives letters for the SQL's data files, log files, etc since
if
two instances failed over to the same failover server the drives would
clash.
Correct? Wow, not sure there are enough drive letters for 6 SQL
instances
using 3-4 drive letters each.
OK - thanks for getting my head on straight.
Darin
"Linchi Shea" wrote:
How could a server not run two clusters in a N+1 configuation.
If I understand this statement correctly, we have some fundamental
problem/misconception here. A node can't be in two different clusters
simultaneously. N+1 simply means that this particular cluster would
have as
many as N+1 nodes, and N of those will be configured to have apps
(i.e. SQL
isntances in this cases) running in the normal operating mode, and
the other
one node will be a spare, to which the other nodes may failover. So,
we
definitely do not have two clusters here.
Linchi
"DTully" wrote:
Well - first TIA!
We have an Active/Passive cluster running W2K3 Enterprise X64 Sp1,
32GB RAM
and SQL Server 2005. We're moving to a N+1 configuration (and then
maybe N+M
later), but I'm running into an error.
After I cluster the third server and try adding the failover node
(it's the
same failover node for the existing cluster), I get the following
error in
the log:
2006-08-30 15:04:30.432 [ERR ] SJCFODBMS016M: This node is a member
of a
different cluster: "SJCADDBMS014V.corp.shop.com". (hr=0x8007000d,
{EBC8AEFF-10C3-4D5B-AC17-FC0F4C3871B7},
{0B5C1D35-1CA5-4597-B1A7-DCD4D64F55C6}, 1, 1, 1), (null)
2006-08-30 15:04:30.495 [ERR ] SJCKWDBMS011V: Checking cluster
membership...
(hr=0x8007000d, {EBC8AEFF-10C3-4D5B-AC17-FC0F4C3871B7},
{06AB21DD-D8C7-453D-845E-50423BEACA38}, 0, 1, 1), (null)
2006-08-30 15:04:30.557 [ERR ] SJCKWDBMS011V: (null)
(hr=0x8007000d,
{FDC75680-7DBB-42CA-8003-E4CDB01FF062},
{EBC8AEFF-10C3-4D5B-AC17-FC0F4C3871B7}, 0, 8, 8), (null)
2006-08-30 15:04:30.604 [INFO] [MT] [CTaskAnalyzeClusterBase]
Exiting task.
The task was not cancelled.
In the wizard I get the same error (that the other node is a member
of a
different cluster) and the install then fails.
I also tried to install the new cluster directly on the failover
server
(cleaned it off the third node first) and it won't even istall and
returns
the same error.
This can't be right. How could a server not run two clusters in a
N+1
configuation. I'm really aiming for N+M with 6 active and 2
dedicated FO
servers. But I'm stuck at the moment.
Thanks and again - TIA!!
Darin
.
- Follow-Ups:
- Re: N+1 Configuration
- From: DTully
- Re: N+1 Configuration
- References:
- RE: N+1 Configuration
- From: DTully
- RE: N+1 Configuration
- From: Linchi Shea
- RE: N+1 Configuration
- From: DTully
- RE: N+1 Configuration
- Prev by Date: RE: N+1 Configuration
- Next by Date: SQL Server 2005 Cluster Install
- Previous by thread: RE: N+1 Configuration
- Next by thread: Re: N+1 Configuration
- Index(es):
Relevant Pages
|
