2-node to 4-node

Hi!
I have a customer with a 2-node active/active sql2k cluster hosted on
w2k.
We now want to run this on w2k3 and we also want to build up the
cluster to have a 4-node active/active/active/active sql2k cluster.
Our plan is to do it in two steps as follow. We take the current 2
nodes offline and unplug them from the network and then install the 2
new nodes with w2k3 using the original ipaddresses and netbiosname and
then restoring the sql2k (one virtuel server and one named instance).
This should give us the same cluster as before but with sql2k running
on w2k3.
This is because the customer will be able to unplug the new nodes and
replug the old ones if there is any trouble with the new nodes.
Then after a few days it's time to reinstall the old nodes with w2k3
and new netbiosnames and connect them to the cluster so we get our
4-node. Then install 2 new named sql instances.
We want all 4 sql instances to be able to run in all 4 nodes.
Now some questions!
When I install a new named instance (on node 3) it will be available
in the other 2 nodes, but will the original sql2k on these 2 nodes be
available in node 3? The same question with node 4.
Is there some problems with sql2k and w2k3 because it is no way to
uninstall sql2k under MSCS? In w2k I'm able to remove it under the
control panel.
/Peter
I actually did a 2 to 4 node cluster upgrade. During the transition, I
replaced the original nodes with new hardware, so I went from 2 nodes with 4
procs each to 4 nodes with 8 procs each. I also went from 2 instances to 3
instances of SQL server. Here is how I did it.
Upgrade the existing nodes (1 at a time) to W2K3. Add the new nodes.
Install SQL and Service packs to the new nodes. Fail the cluster over to
the new nodes. Remove the old nodes and replace with new hardware. At this
point, I would drop those nodes and rebuild them from scratch. Upgrad for
2003 is good, but I prefer a clean build. Add the new nodes to SQL and
apply the service packs. Install the third SQL instance and service pack.
As for your two questions, whenever you add or remove a node, you will need
to use the SQL install CD to add or remove the node from the SQL
installation. This is in addition to evicting or adding the node from the
cluster tool. Choose Advanced Options and Maintain Failover Clustering.
Read BOL on failover clustering for more details.
Geoff N. Hiten
Microsoft SQL Server MVP
Senior Database Administrator
Careerbuilder.com
I support the Professional Association for SQL Server
www.sqlpass.org
"Peter Lindberg" <peter.lindbergTAKE@.AWAYmartinsson.se> wrote in message
news:40641e4b.73299028@.msnews.microsoft.com...
> Hi!
> I have a customer with a 2-node active/active sql2k cluster hosted on
> w2k.
> We now want to run this on w2k3 and we also want to build up the
> cluster to have a 4-node active/active/active/active sql2k cluster.
> Our plan is to do it in two steps as follow. We take the current 2
> nodes offline and unplug them from the network and then install the 2
> new nodes with w2k3 using the original ipaddresses and netbiosname and
> then restoring the sql2k (one virtuel server and one named instance).
> This should give us the same cluster as before but with sql2k running
> on w2k3.
> This is because the customer will be able to unplug the new nodes and
> replug the old ones if there is any trouble with the new nodes.
> Then after a few days it's time to reinstall the old nodes with w2k3
> and new netbiosnames and connect them to the cluster so we get our
> 4-node. Then install 2 new named sql instances.
> We want all 4 sql instances to be able to run in all 4 nodes.
> Now some questions!
> When I install a new named instance (on node 3) it will be available
> in the other 2 nodes, but will the original sql2k on these 2 nodes be
> available in node 3? The same question with node 4.
> Is there some problems with sql2k and w2k3 because it is no way to
> uninstall sql2k under MSCS? In w2k I'm able to remove it under the
> control panel.
> /Peter
>
|||Thank's for your advice Geoff.
The questions comes after a test install I have done and there was a
few odd things that happend.
We use Veritas Volume Manager to allow dynamic disc. I'm not involved
in that part, but the storage techi say's that the veritas code for
w2k can't handle an upgrade of windows to w2k3 and veritas w2k code
can't be upgraded to w2k3 version and still running on w2k. That's why
I want to install w2k3 on new hw with the new codepack from Verirtas.
In our test we run into the following situation:
On node one where we installed the veritas code we connected to disk
and then set up the cluster and then the first sql virtual server. At
that time the other node didn't have the veritas code yet, it only had
w2k3 and was not part of the ckuster. That means all disk was
available to node one. That node was installed with a named instance
where I by the way had to use instruction from KB815431 and sp3a. That
was after the install of sql2k failed once.
After this we set up the veritas code on the other machine with
clustersupport and added it to the cluster. We made an error labeling
one disk with the same name as the other node(cut and paste). When
changing that we had to uninstall the veritas code and break the
cluster. On with veritas code again without clustersupport so we get a
possiblity to chage that label. Then get the node in the cluster
again.
Here I installed a new virtuel server with a new named instance and
sp3a on the second node. This setup was without problem and the only
odd thing was that the dtc service still was automatic.
Next step was to se that the instances could move between the nodes.
The second node could move to the first, but the first could not move
to the second. We found out that the first node have the service for
both instances but the second only the service for the last installed
instance.
I can't se why this happend.
This is why I'm conserned about adding 2 more nodes after a week. Will
the initial two be able to move to new once? If the problem I
described is created by some of my mistake I will still will do this
upgrade in 2 steps, but if there is this problem we have to reinstall
the old nodes to w2k3 at the same time.
I will try to recreate it in vmware workstation, but if someone know
anything about this I will be happy to hear it.
/Peter
On Fri, 26 Mar 2004 10:26:22 -0500, "Geoff N. Hiten"
<SRDBA@.Careerbuilder.com> wrotc:

>I actually did a 2 to 4 node cluster upgrade. During the transition, I
>replaced the original nodes with new hardware, so I went from 2 nodes with 4
>procs each to 4 nodes with 8 procs each. I also went from 2 instances to 3
>instances of SQL server. Here is how I did it.
>Upgrade the existing nodes (1 at a time) to W2K3. Add the new nodes.
>Install SQL and Service packs to the new nodes. Fail the cluster over to
>the new nodes. Remove the old nodes and replace with new hardware. At this
>point, I would drop those nodes and rebuild them from scratch. Upgrad for
>2003 is good, but I prefer a clean build. Add the new nodes to SQL and
>apply the service packs. Install the third SQL instance and service pack.
>As for your two questions, whenever you add or remove a node, you will need
>to use the SQL install CD to add or remove the node from the SQL
>installation. This is in addition to evicting or adding the node from the
>cluster tool. Choose Advanced Options and Maintain Failover Clustering.
>Read BOL on failover clustering for more details.
>--
>Geoff N. Hiten
>Microsoft SQL Server MVP
>Senior Database Administrator
>Careerbuilder.com
>I support the Professional Association for SQL Server
>www.sqlpass.org
>"Peter Lindberg" <peter.lindbergTAKE@.AWAYmartinsson.se> wrote in message
>news:40641e4b.73299028@.msnews.microsoft.com...
>
|||You can determine which nodes can host which instances. Using the Install
CD and the Advanced Options tab, you can add or remove individual nodes to a
particular instance. I added each instance to each new node individually.
I then added the service pack to the new node/instance. Of course, I
rebooted each new node after each step, so there are a lot of reboots.
There is a new feature in W2k3 where you can expand basic cluster disks
without converting them to dynamic disks. That may eliminate the need for
Veritas Volume Manager.
Geoff N. Hiten
Microsoft SQL Server MVP
Senior Database Administrator
Careerbuilder.com
I support the Professional Association for SQL Server
www.sqlpass.org
"Peter Lindberg" <peter.lindberg@.unigrid.se> wrote in message
news:406873b9.247543388@.msnews.microsoft.com...
> Thank's for your advice Geoff.
> The questions comes after a test install I have done and there was a
> few odd things that happend.
> We use Veritas Volume Manager to allow dynamic disc. I'm not involved
> in that part, but the storage techi say's that the veritas code for
> w2k can't handle an upgrade of windows to w2k3 and veritas w2k code
> can't be upgraded to w2k3 version and still running on w2k. That's why
> I want to install w2k3 on new hw with the new codepack from Verirtas.
> In our test we run into the following situation:
> On node one where we installed the veritas code we connected to disk
> and then set up the cluster and then the first sql virtual server. At
> that time the other node didn't have the veritas code yet, it only had
> w2k3 and was not part of the ckuster. That means all disk was
> available to node one. That node was installed with a named instance
> where I by the way had to use instruction from KB815431 and sp3a. That
> was after the install of sql2k failed once.
> After this we set up the veritas code on the other machine with
> clustersupport and added it to the cluster. We made an error labeling
> one disk with the same name as the other node(cut and paste). When
> changing that we had to uninstall the veritas code and break the
> cluster. On with veritas code again without clustersupport so we get a
> possiblity to chage that label. Then get the node in the cluster
> again.
> Here I installed a new virtuel server with a new named instance and
> sp3a on the second node. This setup was without problem and the only
> odd thing was that the dtc service still was automatic.
> Next step was to se that the instances could move between the nodes.
> The second node could move to the first, but the first could not move
> to the second. We found out that the first node have the service for
> both instances but the second only the service for the last installed
> instance.
> I can't se why this happend.
> This is why I'm conserned about adding 2 more nodes after a week. Will
> the initial two be able to move to new once? If the problem I
> described is created by some of my mistake I will still will do this
> upgrade in 2 steps, but if there is this problem we have to reinstall
> the old nodes to w2k3 at the same time.
> I will try to recreate it in vmware workstation, but if someone know
> anything about this I will be happy to hear it.
> /Peter
>
> On Fri, 26 Mar 2004 10:26:22 -0500, "Geoff N. Hiten"
> <SRDBA@.Careerbuilder.com> wrotc:
with 4
3
this
for
pack.
need
the
>
|||I saw that under Advanced Options but the button was greymarked. It
must have happend something. Now I have done it i vmware and seen it
work.
Next will be the actual go later this week, and then two more nodes a
week later. I'll remember those reboots after each part. I get it that
before starting with node 4 you where up with SP3a on node 3.
/Peter
On Mon, 29 Mar 2004 09:39:03 -0500, "Geoff N. Hiten"
<SRDBA@.Careerbuilder.com> wrotc:

>You can determine which nodes can host which instances. Using the Install
>CD and the Advanced Options tab, you can add or remove individual nodes to a
>particular instance. I added each instance to each new node individually.
>I then added the service pack to the new node/instance. Of course, I
>rebooted each new node after each step, so there are a lot of reboots.
>There is a new feature in W2k3 where you can expand basic cluster disks
>without converting them to dynamic disks. That may eliminate the need for
>Veritas Volume Manager.

2005 FT - Unable to re-populate catalogue

Hi,
I have a clustered 2005 server (build 9.00.1399.06). On the server I have 3
FT catalogues.
When I created the first catalogue everything was fine. The catalogue built
fine and returns results correctly (sooo much quicker than 2000!). Then I
tried to do a full re-population.
The full text service appeared to hang and never re-populate the catalogue.
From various bits of testing I have found this:
Using sp_who and sp_lock I can see that the Gatherer spid is blocked by the
FT Monitor spid. The Gatherer has await resource of MD 24(7:0:0) which the
monitor has Sch-S lock on. Being a system process I cannot kill it. During
this time I cannot access the Full text screen in Man Studio (blocked again),
but I can query other catalogues.
When it appears to hang as above I tried stopping the FT service in cluster
admin. The process would not stop. I have to kill it using task manager.
After a restart of just the full text service I cannot access the FT
catalogue screen, though strangley I can't see any blocking. The catalogue
appears not to be rebuilt. I can query other catalogues fine. I cannot see
any blocking of gatherer (I can't see it in sp_who), or any other blocking
spids for that matter.
Last resort I restart the Full text service and SQL Service. Upon restarting
the catalogue starts repopulating straight away. I can access it in Man
Studio, it completes fine and everything is good. Until I try to start a full
population again. (BTW I've tried incremental and rebuilding - same as above
happens).
So in other words I have to restart the SQL service everytime I want to
re-populate my FT catalogue.
I'm hoping someone can help me. Any form of suggestion welcomed!
Thanks
I just found out that FT keeps a log. Here is part of todays output (it is
repeated several times):
2005-11-17 11:16:06.30 spid13s Warning: Full-text population for table
or indexed view '[IMS].[dbo].[MyTable]' failed to send batch of data to
MSFTESQL service (table or indexed view ID '1253579504', catalog ID '5',
database ID '6'). Error: 0x8001010e(The application called an interface that
was marshalled for a different thread.).
2005-11-17 11:16:16.30 spid12s The full-text catalog health monitor
reported a failure for full-text catalog "FTC_MYCatName" (5) in database
"IMS" (6). Reason code: 0. Error: 0x8001010e(The application called an
interface that was marshalled for a different thread.). The system will
restart any in-progress population from the previous checkpoint. If this
message occurs frequently, consult SQL Server Books Online for
troubleshooting assistance. This is an informational message only. No user
action is required.
2005-11-17 11:22:16.86 spid13s Informational: Reinitialized full-text
Full population for table or indexed view '[IMS].[dbo].[MyTable]' (table or
indexed view ID '1253579504', database ID '6') after a temporary failure.
Prior to error, the number of documents processed: 60800, errors encountered:
0.
"Simon" wrote:

> Hi,
> I have a clustered 2005 server (build 9.00.1399.06). On the server I have 3
> FT catalogues.
> When I created the first catalogue everything was fine. The catalogue built
> fine and returns results correctly (sooo much quicker than 2000!). Then I
> tried to do a full re-population.
> The full text service appeared to hang and never re-populate the catalogue.
> From various bits of testing I have found this:
> Using sp_who and sp_lock I can see that the Gatherer spid is blocked by the
> FT Monitor spid. The Gatherer has await resource of MD 24(7:0:0) which the
> monitor has Sch-S lock on. Being a system process I cannot kill it. During
> this time I cannot access the Full text screen in Man Studio (blocked again),
> but I can query other catalogues.
> When it appears to hang as above I tried stopping the FT service in cluster
> admin. The process would not stop. I have to kill it using task manager.
> After a restart of just the full text service I cannot access the FT
> catalogue screen, though strangley I can't see any blocking. The catalogue
> appears not to be rebuilt. I can query other catalogues fine. I cannot see
> any blocking of gatherer (I can't see it in sp_who), or any other blocking
> spids for that matter.
> Last resort I restart the Full text service and SQL Service. Upon restarting
> the catalogue starts repopulating straight away. I can access it in Man
> Studio, it completes fine and everything is good. Until I try to start a full
> population again. (BTW I've tried incremental and rebuilding - same as above
> happens).
> So in other words I have to restart the SQL service everytime I want to
> re-populate my FT catalogue.
> I'm hoping someone can help me. Any form of suggestion welcomed!
> Thanks
>
>
|||Sounds like Monitor is locking with the population process. The population
does put locks on the rows and table it is indexing to preserve
transactional consistency.
It looks like you have to do a less invasive form of monitoring.
Hilary Cotter
Looking for a SQL Server replication book?
http://www.nwsu.com/0974973602.html
Looking for a FAQ on Indexing Services/SQL FTS
http://www.indexserverfaq.com
"Simon" <Simon@.discussions.microsoft.com> wrote in message
news:3B782A40-FF11-4694-B21A-0EFFBA0F5F02@.microsoft.com...[vbcol=seagreen]
>I just found out that FT keeps a log. Here is part of todays output (it is
> repeated several times):
> 2005-11-17 11:16:06.30 spid13s Warning: Full-text population for table
> or indexed view '[IMS].[dbo].[MyTable]' failed to send batch of data to
> MSFTESQL service (table or indexed view ID '1253579504', catalog ID '5',
> database ID '6'). Error: 0x8001010e(The application called an interface
> that
> was marshalled for a different thread.).
> 2005-11-17 11:16:16.30 spid12s The full-text catalog health monitor
> reported a failure for full-text catalog "FTC_MYCatName" (5) in database
> "IMS" (6). Reason code: 0. Error: 0x8001010e(The application called an
> interface that was marshalled for a different thread.). The system will
> restart any in-progress population from the previous checkpoint. If this
> message occurs frequently, consult SQL Server Books Online for
> troubleshooting assistance. This is an informational message only. No user
> action is required.
> 2005-11-17 11:22:16.86 spid13s Informational: Reinitialized full-text
> Full population for table or indexed view '[IMS].[dbo].[MyTable]' (table
> or
> indexed view ID '1253579504', database ID '6') after a temporary failure.
> Prior to error, the number of documents processed: 60800, errors
> encountered:
> 0.
> "Simon" wrote:
|||Hilary,
Thanks for your reply. I agree with the cause. However the FT Monitor
process is a system process not something I have put in place!!
I don't what it is or what it does but I can't do anything to stop it. Do
you know what this process is?
Thanks
"Hilary Cotter" wrote:

> Sounds like Monitor is locking with the population process. The population
> does put locks on the rows and table it is indexing to preserve
> transactional consistency.
> It looks like you have to do a less invasive form of monitoring.
> --
> Hilary Cotter
> Looking for a SQL Server replication book?
> http://www.nwsu.com/0974973602.html
> Looking for a FAQ on Indexing Services/SQL FTS
> http://www.indexserverfaq.com
> "Simon" <Simon@.discussions.microsoft.com> wrote in message
> news:3B782A40-FF11-4694-B21A-0EFFBA0F5F02@.microsoft.com...
>
>
|||Is anyone able to help on this?
In short I'm trying to resolve this error:
2005-11-17 11:16:06.30 spid13s Warning: Full-text population for
table
or indexed view '[IMS].[dbo].[MyTable]' failed to send batch of data to
MSFTESQL service (table or indexed view ID '1253579504', catalog ID
'5',
database ID '6'). Error: 0x8001010e(The application called an interface
that
was marshalled for a different thread.).
Which occurs when I populate my catalogue.
Thanks
Simon

2005 Express Edition CREATE statement issues

Hello everyone,

i am using the Express Edition to build an application which requires rather large databases to store binary documents. I've been experiencing weird behaviour with the Express Edition on Vista and XP running the following CREATE statement:

Code Snippet

USE master;
GO
CREATE DATABASE [document_archive]
ON PRIMARY
( NAME='document_archive_Primary',
FILENAME='D:\Data\document_archive.mdf',
SIZE=32MB,
MAXSIZE=1536MB,
FILEGROWTH=10%)
LOG ON
( NAME='document_archive_Log',
FILENAME = 'D:\Data\document_archive.ldf',
SIZE=8MB,
MAXSIZE=64MB,
FILEGROWTH=10%);
GO

The server is telling me that everything went ok but the database doesent show up in the TreeView, nor it is accessible. Retrying to execute the script i cought an error telling me that the database couldnt be created because of existing files, which is the behaviour Id expect.

I dropped the database and went testing the script on a SQL 2000 Server Enterprise Edition which worked as expected, immediately showing up in the TreeView.

As far i remember the 2005 Express Edition can handle databases up to 2GB (or maybe 5GB? Dont remember exactly).

I am having a hard time to figure out what might cause the weird behaviour.

However, heres the system specs of the two 2005 Express installations:

Box#1: Vista RC1

Box#2: XP SP 2

Both running SQL Server 2005 Express Edition SP2, installed on 'C:\Program Files\Microsoft SQL Server', having tons of free disk space and spare resources to handle this CREATE statement and requests.

Any hint that might help solve my issue would be greatly appreciated

Many thanks in advance,

Yours,

trahho.

How are running this script?

When you run the script above, can you immediately run a select * from sys.databases and see if your database is listed.

The DB limit for Express is 4gb so there is not a problem there.