Running Sql Server 2000 SP3a on a two node cluster with one instance
running on the active node (Active/Passive)
We have to replace one of the physical nodes (repeated hw failures) -
how do I go about replacing the node within SQL Server. The NT guys
are telling me - that have to install from scratch the OS and bring it
up to the same OS patch levels of the remaining stable node.
So, once I the new node has been re-added to the phsical cluster...how
do i get SQL Server 2000 installed and how do I get it upgraded to SPS
3A
Joe E O
Surprising that you should ask that. On 10/27, I posted the following list
of instructions.
Good luck.
Anthony Thomas
It is easy to evict a cluster node, repair or rebuild, and, perhaps even
upgrade it; however, on many occasions, we have had considerable trouble
reintroducing that newly built node back into the cluster configuration.
The following represents the cleanest solution we have come up with to date:
Here's the sequence of events.
1. Run SQL Server setup to manage the virtual server configuration and add
the new node to this configuration. This installs the RTM binaries to the
new node. It also installs the dreaded SCM (Service Control Manager), but
also the Cluster Network utility. (KB295589)
2. Log into the new node and remove the SCM from the system tray and startup
group. (Service Pack Read Me docs).
3. On both nodes, create Client Side Aliases using Named Pipe for the
virtual server instance. (KB815431)
4. Reboot the added node.
5. Apply SP3a (this was the prior headache). If you do not, because of the
3-digit Build number versus the SP4 4-digit build number, that stupid Win2K3
security dialog about only supporting SS2K if SP3 or later will launch on
the
unattended remote installation in interactive mode, which hangs. (KB905286)
(KB902955) (KB329329)
6. Reboot the added node.
7. Apply SP4.
8. Reboot the added node.
9. Apply HF 2187.
10. Reboot the added node.
11. On both nodes, switch the Client Side aliases to TCP protocol.
12. Now, here's the new error. When attempting to run the instance on the
newly added node, the cluster log will report that it could not start the
sqlsrvres and fail with error 435.
What this means is that the registry still contains the default out of box
registry keys. Primarily, the Named Pipe default pipe and the TCP default
port, as well as the default Start Up parameters. If any of these have been
modified from default, services will start up but fail on the newly added
node.
The fix is to log into the new node and use regedit to connect to the other
node. Export the following keys and save them to the local node.
HKLM\Cluster
HKLM\SOFTWARE\Microsoft\Microsoft SQL Server
HKLM\SOFTWARE\Microsoft\MSSQLSERVER
Then merge these exports to the new node (local registry).
This last one is undocumented. If anyone kind find a KB reference, it would
be appreciated.
Sincerely,
Anthony Thomas
"Joe E O" <josephobrien@.hotmail.com> wrote in message
news:1163777568.605425.238240@.j44g2000cwa.googlegr oups.com...
> Running Sql Server 2000 SP3a on a two node cluster with one instance
> running on the active node (Active/Passive)
> We have to replace one of the physical nodes (repeated hw failures) -
> how do I go about replacing the node within SQL Server. The NT guys
> are telling me - that have to install from scratch the OS and bring it
> up to the same OS patch levels of the remaining stable node.
> So, once I the new node has been re-added to the phsical cluster...how
> do i get SQL Server 2000 installed and how do I get it upgraded to SPS
> 3A
>
> Joe E O
>
sql
Showing posts with label passive. Show all posts
Showing posts with label passive. Show all posts
Friday, March 30, 2012
Monday, March 26, 2012
Read Only Reporting Server Active/Active Configuration
I currently have an active/passive cluster acting as our OLTP TX database
server that is replicating to another active/passive cluster acting as a
Reporting database server.
I have heard that if the database is read-only (e.g. reporting server) that
it is possible to set the sql server application in an active/active mode and
have requests be handled by both of the servers (e.g. shared access to a
single databse). This is what I want to do, but I don't think it is possible
since one node will be given access to the disk resources, correct?
I guess the only way to really use active/active is to have two sql server
instances each with its own disk resources that are supporting two
independent applications and then upon failure the non-failed node handles
both applications. This isn't really what I want to do because I want to
load-balance read-only requests to the same database across two servers;
which I am not able to do. Am I understanding this correctly?
The correct name is scalable shared databases, where multiple SQL 2005
servers are accessing a read-only volume.
there are many restrictions to this
Please see ; http://support.microsoft.com/default...b;en-us;910378
for more info on this
HTH,
_Edwin.
"Larry Herbinaux" <LarryHerbinaux@.discussions.microsoft.com> wrote in
message news:981C4446-DC47-43D3-B071-742BCEE767F3@.microsoft.com...
> I currently have an active/passive cluster acting as our OLTP TX database
> server that is replicating to another active/passive cluster acting as a
> Reporting database server.
> I have heard that if the database is read-only (e.g. reporting server)
that
> it is possible to set the sql server application in an active/active mode
and
> have requests be handled by both of the servers (e.g. shared access to a
> single databse). This is what I want to do, but I don't think it is
possible
> since one node will be given access to the disk resources, correct?
> I guess the only way to really use active/active is to have two sql server
> instances each with its own disk resources that are supporting two
> independent applications and then upon failure the non-failed node handles
> both applications. This isn't really what I want to do because I want to
> load-balance read-only requests to the same database across two servers;
> which I am not able to do. Am I understanding this correctly?
>
|||Thanks Edwin,
Is this also supported for SQL 2000 or just SQL 2005?
If reporting really needed to scale and the reporting requirements for some
reports needed to be up to the minute but most of the reports could use data
that was a day old, then this seems like the following configuration would
seem to give you the most bang for your buck, correct?
TX Database - 2-Node Active/Passive
Reporting Database Replicated - 2-Node Active/Passive - This would have the
entire TX Database replicated in real-time like we have today. Client access
to this database would be limited to the reports that needed to be accurate
to the minute.
Reporting Database Manual - N-Node Active/Active/Active... - This would have
two volumes as discussed in the article, and the operational volume would be
swapped out with the newly built volume just after midnight with a newer data
set. Client access to this database would limited to the reports that don't
need to be up to the minute accurate. During the downtime, access to the
Reporting Database Replicated cluster could be changed to allow all reports
to be run from it; given the fact that these changes would be done at
midnight, the load on this servers shouldn't be very great so it should be
able to handle the additional reports.
"Edwin vMierlo" wrote:
> The correct name is scalable shared databases, where multiple SQL 2005
> servers are accessing a read-only volume.
> there are many restrictions to this
> Please see ; http://support.microsoft.com/default...b;en-us;910378
> for more info on this
> HTH,
> _Edwin.
>
> "Larry Herbinaux" <LarryHerbinaux@.discussions.microsoft.com> wrote in
> message news:981C4446-DC47-43D3-B071-742BCEE767F3@.microsoft.com...
> that
> and
> possible
>
>
|||answers inline
"Larry Herbinaux" <LarryHerbinaux@.discussions.microsoft.com> wrote in
message news:F5FD9768-8D48-409C-8431-338CFD7B9D49@.microsoft.com...
> Thanks Edwin,
> Is this also supported for SQL 2000 or just SQL 2005?
SQL 2005 Enterprise Edition
> If reporting really needed to scale and the reporting requirements for
some
> reports needed to be up to the minute but most of the reports could use
data
> that was a day old, then this seems like the following configuration would
> seem to give you the most bang for your buck, correct?
> TX Database - 2-Node Active/Passive
> Reporting Database Replicated - 2-Node Active/Passive - This would have
the
> entire TX Database replicated in real-time like we have today. Client
access
> to this database would be limited to the reports that needed to be
accurate
> to the minute.
> Reporting Database Manual - N-Node Active/Active/Active... - This would
have
> two volumes as discussed in the article, and the operational volume would
be
> swapped out with the newly built volume just after midnight with a newer
data
> set. Client access to this database would limited to the reports that
don't
> need to be up to the minute accurate. During the downtime, access to the
> Reporting Database Replicated cluster could be changed to allow all
reports
> to be run from it; given the fact that these changes would be done at
> midnight, the load on this servers shouldn't be very great so it should be
> able to handle the additional reports.
Yes, while you are "swapping" in the new data, you will have downtime
[vbcol=seagreen]
> "Edwin vMierlo" wrote:
http://support.microsoft.com/default...b;en-us;910378[vbcol=seagreen]
database[vbcol=seagreen]
a[vbcol=seagreen]
mode[vbcol=seagreen]
a[vbcol=seagreen]
server[vbcol=seagreen]
handles[vbcol=seagreen]
to[vbcol=seagreen]
servers;[vbcol=seagreen]
server that is replicating to another active/passive cluster acting as a
Reporting database server.
I have heard that if the database is read-only (e.g. reporting server) that
it is possible to set the sql server application in an active/active mode and
have requests be handled by both of the servers (e.g. shared access to a
single databse). This is what I want to do, but I don't think it is possible
since one node will be given access to the disk resources, correct?
I guess the only way to really use active/active is to have two sql server
instances each with its own disk resources that are supporting two
independent applications and then upon failure the non-failed node handles
both applications. This isn't really what I want to do because I want to
load-balance read-only requests to the same database across two servers;
which I am not able to do. Am I understanding this correctly?
The correct name is scalable shared databases, where multiple SQL 2005
servers are accessing a read-only volume.
there are many restrictions to this
Please see ; http://support.microsoft.com/default...b;en-us;910378
for more info on this
HTH,
_Edwin.
"Larry Herbinaux" <LarryHerbinaux@.discussions.microsoft.com> wrote in
message news:981C4446-DC47-43D3-B071-742BCEE767F3@.microsoft.com...
> I currently have an active/passive cluster acting as our OLTP TX database
> server that is replicating to another active/passive cluster acting as a
> Reporting database server.
> I have heard that if the database is read-only (e.g. reporting server)
that
> it is possible to set the sql server application in an active/active mode
and
> have requests be handled by both of the servers (e.g. shared access to a
> single databse). This is what I want to do, but I don't think it is
possible
> since one node will be given access to the disk resources, correct?
> I guess the only way to really use active/active is to have two sql server
> instances each with its own disk resources that are supporting two
> independent applications and then upon failure the non-failed node handles
> both applications. This isn't really what I want to do because I want to
> load-balance read-only requests to the same database across two servers;
> which I am not able to do. Am I understanding this correctly?
>
|||Thanks Edwin,
Is this also supported for SQL 2000 or just SQL 2005?
If reporting really needed to scale and the reporting requirements for some
reports needed to be up to the minute but most of the reports could use data
that was a day old, then this seems like the following configuration would
seem to give you the most bang for your buck, correct?
TX Database - 2-Node Active/Passive
Reporting Database Replicated - 2-Node Active/Passive - This would have the
entire TX Database replicated in real-time like we have today. Client access
to this database would be limited to the reports that needed to be accurate
to the minute.
Reporting Database Manual - N-Node Active/Active/Active... - This would have
two volumes as discussed in the article, and the operational volume would be
swapped out with the newly built volume just after midnight with a newer data
set. Client access to this database would limited to the reports that don't
need to be up to the minute accurate. During the downtime, access to the
Reporting Database Replicated cluster could be changed to allow all reports
to be run from it; given the fact that these changes would be done at
midnight, the load on this servers shouldn't be very great so it should be
able to handle the additional reports.
"Edwin vMierlo" wrote:
> The correct name is scalable shared databases, where multiple SQL 2005
> servers are accessing a read-only volume.
> there are many restrictions to this
> Please see ; http://support.microsoft.com/default...b;en-us;910378
> for more info on this
> HTH,
> _Edwin.
>
> "Larry Herbinaux" <LarryHerbinaux@.discussions.microsoft.com> wrote in
> message news:981C4446-DC47-43D3-B071-742BCEE767F3@.microsoft.com...
> that
> and
> possible
>
>
|||answers inline
"Larry Herbinaux" <LarryHerbinaux@.discussions.microsoft.com> wrote in
message news:F5FD9768-8D48-409C-8431-338CFD7B9D49@.microsoft.com...
> Thanks Edwin,
> Is this also supported for SQL 2000 or just SQL 2005?
SQL 2005 Enterprise Edition
> If reporting really needed to scale and the reporting requirements for
some
> reports needed to be up to the minute but most of the reports could use
data
> that was a day old, then this seems like the following configuration would
> seem to give you the most bang for your buck, correct?
> TX Database - 2-Node Active/Passive
> Reporting Database Replicated - 2-Node Active/Passive - This would have
the
> entire TX Database replicated in real-time like we have today. Client
access
> to this database would be limited to the reports that needed to be
accurate
> to the minute.
> Reporting Database Manual - N-Node Active/Active/Active... - This would
have
> two volumes as discussed in the article, and the operational volume would
be
> swapped out with the newly built volume just after midnight with a newer
data
> set. Client access to this database would limited to the reports that
don't
> need to be up to the minute accurate. During the downtime, access to the
> Reporting Database Replicated cluster could be changed to allow all
reports
> to be run from it; given the fact that these changes would be done at
> midnight, the load on this servers shouldn't be very great so it should be
> able to handle the additional reports.
Yes, while you are "swapping" in the new data, you will have downtime
[vbcol=seagreen]
> "Edwin vMierlo" wrote:
http://support.microsoft.com/default...b;en-us;910378[vbcol=seagreen]
database[vbcol=seagreen]
a[vbcol=seagreen]
mode[vbcol=seagreen]
a[vbcol=seagreen]
server[vbcol=seagreen]
handles[vbcol=seagreen]
to[vbcol=seagreen]
servers;[vbcol=seagreen]
Labels:
acting,
active,
cluster,
configuration,
database,
databaseserver,
microsoft,
mysql,
oltp,
oracle,
passive,
replicating,
reporting,
server,
sql
Subscribe to:
Posts (Atom)