Optimal CF Server settings w/SQL Server Cluster

2003-12-01 Thread Peter Amiri
I have a few questions regarding server settings in the following environement:

CF 5.0 cluster with 5 servers in the farm
MS SQL Server cluster - Raid 10 array connected to both servers with hot fail over
The application gets about 7 million page impressions per day 

I want to what the optimal settings should be for data sources settings. In particular Maintain Connections and Timeout settings. Currently maintain connections is set to yes and timeout is set to 5 minutes. But when the SQL server locks up for whatever reason and fails over to the backup server the CF queries just start to queue up. In other words it apears that maintaining connections open by CF some how does not get updated to the newly live box. Before resorting to turing off maintain connections I wanted to know if anyone has had much success with CF and SQL clusters.

-Peter Amiri
 [EMAIL PROTECTED]
 MySpace.com
 [Todays Threads] 
 [This Message] 
 [Subscription] 
 [Fast Unsubscribe] 
 [User Settings]




RE: Optimal CF Server settings w/SQL Server Cluster

2003-12-01 Thread Stacy Young
I'm no expert in this area but if by chance you have switching
capabilities in your topology (Alteon or what-not)...create a vlan for
the database cluster and expose that to the CF servers via a single
network address. That way CF is always bound to a static virtual server
address...no matter which database cluster member is operating behind
the scenes.

Stace

_

From: Peter Amiri [mailto:[EMAIL PROTECTED] 
Sent: December 1, 2003 2:30 PM
To: CF-Talk
Subject: Optimal CF Server settings w/SQL Server Cluster

I have a few questions regarding server settings in the following
environement:

CF 5.0 cluster with 5 servers in the farm
MS SQL Server cluster - Raid 10 array connected to both servers with hot
fail over
The application gets about 7 million page impressions per day 

I want to what the optimal settings should be for data sources settings.
In particular Maintain Connections and Timeout settings. Currently
maintain connections is set to yes and timeout is set to 5 minutes. But
when the SQL server locks up for whatever reason and fails over to the
backup server the CF queries just start to queue up. In other words it
apears that maintaining connections open by CF some how does not get
updated to the newly live box. Before resorting to turing off maintain
connections I wanted to know if anyone has had much success with CF and
SQL clusters.

-Peter Amiri
[EMAIL PROTECTED]
MySpace.com

_
 [Todays Threads] 
 [This Message] 
 [Subscription] 
 [Fast Unsubscribe] 
 [User Settings]




RE: Optimal CF Server settings w/SQL Server Cluster

2003-12-01 Thread Peter Amiri
Stace,

 
It's my understanding the the MS SQL Server cluster does this for you. Althought each of the individual servers have their own set of IP addresses. A virtual SQL Server is also created with it's own ip. All the CF Servers are pointed to the virtual SQL Server. When a SQL Server fail over occurs the CF Servers continue to communicate to the database through the same ip addresss. So therefore a seperate VLAN to achive the same thing shouldn't be needed. I may be wrong though? I suspect if I wait for the five minute timeout to occur the old connections will get killed and new ones will be created which would then work without a hitch. I can't afford the 5 minutes of dead time though. I am wondering how much of a savings Maintain Connections provides and whether we should just nix that options which should solve the problem.

-Peter Amiri 
 [EMAIL PROTECTED] 
 MySpace.com 

-Original Message-
From: Stacy Young [mailto:[EMAIL PROTECTED]
Sent: Monday, December 01, 2003 11:35 AM
To: CF-Talk
Subject: RE: Optimal CF Server settings w/SQL Server Cluster

I'm no expert in this area but if by chance you have switching
capabilities in your topology (Alteon or what-not)...create a vlan for
the database cluster and expose that to the CF servers via a single
network address. That way CF is always bound to a static virtual server
address...no matter which database cluster member is operating behind
the scenes.

Stace

_

From: Peter Amiri [mailto:[EMAIL PROTECTED] 
Sent: December 1, 2003 2:30 PM
To: CF-Talk
Subject: Optimal CF Server settings w/SQL Server Cluster

I have a few questions regarding server settings in the following
environement:

CF 5.0 cluster with 5 servers in the farm
MS SQL Server cluster - Raid 10 array connected to both servers with hot
fail over
The application gets about 7 million page impressions per day 

I want to what the optimal settings should be for data sources settings.
In particular Maintain Connections and Timeout settings. Currently
maintain connections is set to yes and timeout is set to 5 minutes. But
when the SQL server locks up for whatever reason and fails over to the
backup server the CF queries just start to queue up. In other words it
apears that maintaining connections open by CF some how does not get
updated to the newly live box. Before resorting to turing off maintain
connections I wanted to know if anyone has had much success with CF and
SQL clusters.

-Peter Amiri
[EMAIL PROTECTED]
MySpace.com

_ 
_
 [Todays Threads] 
 [This Message] 
 [Subscription] 
 [Fast Unsubscribe] 
 [User Settings]




RE: Optimal CF Server settings w/SQL Server Cluster

2003-12-01 Thread Dave Watts
 It's my understanding the the MS SQL Server cluster does this 
 for you. Althought each of the individual servers have their 
 own set of IP addresses. A virtual SQL Server is also created 
 with it's own ip. All the CF Servers are pointed to the 
 virtual SQL Server. When a SQL Server fail over occurs the CF 
 Servers continue to communicate to the database through the 
 same ip addresss. So therefore a seperate VLAN to achive the 
 same thing shouldn't be needed. I may be wrong though?

No, I think you're right, if your SQL Server cluster is set up as
active-active. You should be addressing the cluster itself, not any
individual server in the cluster.

 I suspect if I wait for the five minute timeout to occur the 
 old connections will get killed and new ones will be created 
 which would then work without a hitch. I can't afford the 5 
 minutes of dead time though. I am wondering how much of a 
 savings Maintain Connections provides and whether we should 
 just nix that options which should solve the problem.

You might consider shortening the time a database connection is maintained.
Maintaining connections does make a pretty big difference with SQL Server,
in my experience. I'd also be concerned if one of your database cluster
members is failing frequently, though - I think that SQL Server's clustering
ability is more useful for load-balancing than it is for failover.

Dave Watts, CTO, Fig Leaf Software
http://www.figleaf.com/
voice: (202) 797-5496
fax: (202) 797-5444
 [Todays Threads] 
 [This Message] 
 [Subscription] 
 [Fast Unsubscribe] 
 [User Settings]




RE: Optimal CF Server settings w/SQL Server Cluster

2003-12-01 Thread Peter Amiri
Dave thanks for the input. My dba tells me that the SQL Server cluster is an active-active so your were right in that aspect. Do you know why the queries would get queued up after a fail-over? Is there any way to force the connections to be killed maybe from the database side to force CF to recreate these?

-Peter Amiri 
 [EMAIL PROTECTED] 
 MySpace.com 

-Original Message-
From: Dave Watts [mailto:[EMAIL PROTECTED]
Sent: Monday, December 01, 2003 12:24 PM
To: CF-Talk
Subject: RE: Optimal CF Server settings w/SQL Server Cluster

 It's my understanding the the MS SQL Server cluster does this 
 for you. Althought each of the individual servers have their 
 own set of IP addresses. A virtual SQL Server is also created 
 with it's own ip. All the CF Servers are pointed to the 
 virtual SQL Server. When a SQL Server fail over occurs the CF 
 Servers continue to communicate to the database through the 
 same ip addresss. So therefore a seperate VLAN to achive the 
 same thing shouldn't be needed. I may be wrong though?

No, I think you're right, if your SQL Server cluster is set up as
active-active. You should be addressing the cluster itself, not any
individual server in the cluster.

 I suspect if I wait for the five minute timeout to occur the 
 old connections will get killed and new ones will be created 
 which would then work without a hitch. I can't afford the 5 
 minutes of dead time though. I am wondering how much of a 
 savings Maintain Connections provides and whether we should 
 just nix that options which should solve the problem.

You might consider shortening the time a database connection is maintained.
Maintaining connections does make a pretty big difference with SQL Server,
in my experience. I'd also be concerned if one of your database cluster
members is failing frequently, though - I think that SQL Server's clustering
ability is more useful for load-balancing than it is for failover.

Dave Watts, CTO, Fig Leaf Software
http://www.figleaf.com/
voice: (202) 797-5496
fax: (202) 797-5444 
_
 [Todays Threads] 
 [This Message] 
 [Subscription] 
 [Fast Unsubscribe] 
 [User Settings]




RE: Optimal CF Server settings w/SQL Server Cluster

2003-12-01 Thread Dave Watts
 Dave thanks for the input. My dba tells me that the SQL 
 Server cluster is an active-active so your were right in that 
 aspect. Do you know why the queries would get queued up after 
 a fail-over? Is there any way to force the connections to be 
 killed maybe from the database side to force CF to recreate these?

Unfortunately, I don't know that much about SQL Server clustering. According
to this MS article, it's up to the client application to close connections:

http://support.microsoft.com/default.aspx?scid=kb;en-us;273673Product=sql2k

Dave Watts, CTO, Fig Leaf Software
http://www.figleaf.com/
voice: (202) 797-5496
fax: (202) 797-5444
 [Todays Threads] 
 [This Message] 
 [Subscription] 
 [Fast Unsubscribe] 
 [User Settings]




RE: Optimal CF Server settings w/SQL Server Cluster

2003-12-01 Thread Peter Amiri
Dave thanks for the link to the article. What I'm going to do is keep the Maintain Connection on and set a 5 minute time out. But I'll create a probe that does a simple DB query and set it to run every minute. If the probe fails, I'll recycle the CF server to clear the database connections. 

 
-Peter Amiri 
 [EMAIL PROTECTED] 
 MySpace.com 

-Original Message-
From: Dave Watts [mailto:[EMAIL PROTECTED]
Sent: Monday, December 01, 2003 1:21 PM
To: CF-Talk
Subject: RE: Optimal CF Server settings w/SQL Server Cluster

 Dave thanks for the input. My dba tells me that the SQL 
 Server cluster is an active-active so your were right in that 
 aspect. Do you know why the queries would get queued up after 
 a fail-over? Is there any way to force the connections to be 
 killed maybe from the database side to force CF to recreate these?

Unfortunately, I don't know that much about SQL Server clustering. According
to this MS article, it's up to the client application to close connections:

http://support.microsoft.com/default.aspx?scid=kb;en-us;273673 http://support.microsoft.com/default.aspx?scid=kb;en-us;273673Product=sql2k Product=sql2k

Dave Watts, CTO, Fig Leaf Software
http://www.figleaf.com/
voice: (202) 797-5496
fax: (202) 797-5444 
_
 [Todays Threads] 
 [This Message] 
 [Subscription] 
 [Fast Unsubscribe] 
 [User Settings]




RE: Optimal CF Server settings w/SQL Server Cluster

2003-12-01 Thread Stacy Young
Ok I gotcha...how about catching the specific exception raised when the
DB does its switchover and manually flush DB connections? 

Stace

_

From: Peter Amiri [mailto:[EMAIL PROTECTED] 
Sent: December 1, 2003 2:54 PM
To: CF-Talk
Subject: RE: Optimal CF Server settings w/SQL Server Cluster

Stace,

It's my understanding the the MS SQL Server cluster does this for you.
Althought each of the individual servers have their own set of IP
addresses. A virtual SQL Server is also created with it's own ip. All
the CF Servers are pointed to the virtual SQL Server. When a SQL Server
fail over occurs the CF Servers continue to communicate to the database
through the same ip addresss. So therefore a seperate VLAN to achive the
same thing shouldn't be needed. I may be wrong though? I suspect if I
wait for the five minute timeout to occur the old connections will get
killed and new ones will be created which would then work without a
hitch. I can't afford the 5 minutes of dead time though. I am wondering
how much of a savings Maintain Connections provides and whether we
should just nix that options which should solve the problem.

-Peter Amiri 
[EMAIL PROTECTED] 
MySpace.com 

-Original Message-
From: Stacy Young [mailto:[EMAIL PROTECTED]
Sent: Monday, December 01, 2003 11:35 AM
To: CF-Talk
Subject: RE: Optimal CF Server settings w/SQL Server Cluster

I'm no expert in this area but if by chance you have switching
capabilities in your topology (Alteon or what-not)...create a vlan for
the database cluster and expose that to the CF servers via a single
network address. That way CF is always bound to a static virtual server
address...no matter which database cluster member is operating behind
the scenes.

Stace

_

From: Peter Amiri [mailto:[EMAIL PROTECTED] 
Sent: December 1, 2003 2:30 PM
To: CF-Talk
Subject: Optimal CF Server settings w/SQL Server Cluster

I have a few questions regarding server settings in the following
environement:

CF 5.0 cluster with 5 servers in the farm
MS SQL Server cluster - Raid 10 array connected to both servers with hot
fail over
The application gets about 7 million page impressions per day 

I want to what the optimal settings should be for data sources settings.
In particular Maintain Connections and Timeout settings. Currently
maintain connections is set to yes and timeout is set to 5 minutes. But
when the SQL server locks up for whatever reason and fails over to the
backup server the CF queries just start to queue up. In other words it
apears that maintaining connections open by CF some how does not get
updated to the newly live box. Before resorting to turing off maintain
connections I wanted to know if anyone has had much success with CF and
SQL clusters.

-Peter Amiri
[EMAIL PROTECTED]
MySpace.com

_ 
_

_
 [Todays Threads] 
 [This Message] 
 [Subscription] 
 [Fast Unsubscribe] 
 [User Settings]




RE: Optimal CF Server settings w/SQL Server Cluster

2003-12-01 Thread Peter Amiri
That is along the line of where I'm going. Do you know how to programmically flush the DB connections. In CF 5 administrator there is a button to flush all DB connections. I haven't seen this in MX administrator. Does this mean there was an unpubished tag to do this and if so is it still there in MX?

-Peter Amiri 
 [EMAIL PROTECTED] 
 MySpace.com 

-Original Message-
From: Stacy Young [mailto:[EMAIL PROTECTED]
Sent: Monday, December 01, 2003 2:43 PM
To: CF-Talk
Subject: RE: Optimal CF Server settings w/SQL Server Cluster

Ok I gotcha...how about catching the specific exception raised when the
DB does its switchover and manually flush DB connections? 

Stace

_

From: Peter Amiri [mailto:[EMAIL PROTECTED] 
Sent: December 1, 2003 2:54 PM
To: CF-Talk
Subject: RE: Optimal CF Server settings w/SQL Server Cluster

Stace,

It's my understanding the the MS SQL Server cluster does this for you.
Althought each of the individual servers have their own set of IP
addresses. A virtual SQL Server is also created with it's own ip. All
the CF Servers are pointed to the virtual SQL Server. When a SQL Server
fail over occurs the CF Servers continue to communicate to the database
through the same ip addresss. So therefore a seperate VLAN to achive the
same thing shouldn't be needed. I may be wrong though? I suspect if I
wait for the five minute timeout to occur the old connections will get
killed and new ones will be created which would then work without a
hitch. I can't afford the 5 minutes of dead time though. I am wondering
how much of a savings Maintain Connections provides and whether we
should just nix that options which should solve the problem.

-Peter Amiri 
[EMAIL PROTECTED] 
MySpace.com 

-Original Message-
From: Stacy Young [mailto:[EMAIL PROTECTED]
Sent: Monday, December 01, 2003 11:35 AM
To: CF-Talk
Subject: RE: Optimal CF Server settings w/SQL Server Cluster

I'm no expert in this area but if by chance you have switching
capabilities in your topology (Alteon or what-not)...create a vlan for
the database cluster and expose that to the CF servers via a single
network address. That way CF is always bound to a static virtual server
address...no matter which database cluster member is operating behind
the scenes.

Stace

_

From: Peter Amiri [mailto:[EMAIL PROTECTED] 
Sent: December 1, 2003 2:30 PM
To: CF-Talk
Subject: Optimal CF Server settings w/SQL Server Cluster

I have a few questions regarding server settings in the following
environement:

CF 5.0 cluster with 5 servers in the farm
MS SQL Server cluster - Raid 10 array connected to both servers with hot
fail over
The application gets about 7 million page impressions per day 

I want to what the optimal settings should be for data sources settings.
In particular Maintain Connections and Timeout settings. Currently
maintain connections is set to yes and timeout is set to 5 minutes. But
when the SQL server locks up for whatever reason and fails over to the
backup server the CF queries just start to queue up. In other words it
apears that maintaining connections open by CF some how does not get
updated to the newly live box. Before resorting to turing off maintain
connections I wanted to know if anyone has had much success with CF and
SQL clusters.

-Peter Amiri
[EMAIL PROTECTED]
MySpace.com

_ 
_

_ 
_
 [Todays Threads] 
 [This Message] 
 [Subscription] 
 [Fast Unsubscribe] 
 [User Settings]




RE: Optimal CF Server settings w/SQL Server Cluster

2003-12-01 Thread Dave Watts
 Ok I gotcha...how about catching the specific exception 
 raised when the DB does its switchover and manually flush 
 DB connections?

I think the problem here is that you don't want to wait for the duration of
a database connection timeout.

Dave Watts, CTO, Fig Leaf Software
http://www.figleaf.com/
voice: (202) 797-5496
fax: (202) 797-5444
 [Todays Threads] 
 [This Message] 
 [Subscription] 
 [Fast Unsubscribe] 
 [User Settings]




RE: Optimal CF Server settings w/SQL Server Cluster

2003-12-01 Thread Dave Watts
 That is along the line of where I'm going. Do you know how to 
 programmically flush the DB connections. In CF 5 administrator 
 there is a button to flush all DB connections. I haven't seen 
 this in MX administrator. Does this mean there was an unpubished 
 tag to do this and if so is it still there in MX?

I think you could do this using the ServiceFactory interface in MX:

http://spike.oli.tudelft.nl/jochemd/index.cfm?PageID=12 (courtesy of Jochem)

Dave Watts, CTO, Fig Leaf Software
http://www.figleaf.com/
voice: (202) 797-5496
fax: (202) 797-5444
 [Todays Threads] 
 [This Message] 
 [Subscription] 
 [Fast Unsubscribe] 
 [User Settings]




RE: Optimal CF Server settings w/SQL Server Cluster

2003-12-01 Thread Peter Amiri
Dave thanks again. I think this will do the trick.

-Peter Amiri 
 [EMAIL PROTECTED] 
 MySpace.com 

-Original Message-
From: Dave Watts [mailto:[EMAIL PROTECTED]
Sent: Monday, December 01, 2003 3:10 PM
To: CF-Talk
Subject: RE: Optimal CF Server settings w/SQL Server Cluster

 That is along the line of where I'm going. Do you know how to 
 programmically flush the DB connections. In CF 5 administrator 
 there is a button to flush all DB connections. I haven't seen 
 this in MX administrator. Does this mean there was an unpubished 
 tag to do this and if so is it still there in MX?

I think you could do this using the ServiceFactory interface in MX:

http://spike.oli.tudelft.nl/jochemd/index.cfm?PageID=12 (courtesy of Jochem)

Dave Watts, CTO, Fig Leaf Software
http://www.figleaf.com/
voice: (202) 797-5496
fax: (202) 797-5444 
_
 [Todays Threads] 
 [This Message] 
 [Subscription] 
 [Fast Unsubscribe] 
 [User Settings]




RE: Optimal CF Server settings w/SQL Server Cluster

2003-12-01 Thread Stacy Young
Weird coincidence but we just ran into this exact problem with an Oracle
cluster. One of the cluster members experienced hardware problems and it
failed over to another cluster member...and CFMX DB connections hung.

Stace

_

From: Dave Watts [mailto:[EMAIL PROTECTED] 
Sent: December 1, 2003 4:21 PM
To: CF-Talk
Subject: RE: Optimal CF Server settings w/SQL Server Cluster

 Dave thanks for the input. My dba tells me that the SQL 
 Server cluster is an active-active so your were right in that 
 aspect. Do you know why the queries would get queued up after 
 a fail-over? Is there any way to force the connections to be 
 killed maybe from the database side to force CF to recreate these?

Unfortunately, I don't know that much about SQL Server clustering.
According
to this MS article, it's up to the client application to close
connections:

http://support.microsoft.com/default.aspx?scid=kb;en-us;273673Product=s
ql2k

Dave Watts, CTO, Fig Leaf Software
http://www.figleaf.com/
voice: (202) 797-5496
fax: (202) 797-5444

_
 [Todays Threads] 
 [This Message] 
 [Subscription] 
 [Fast Unsubscribe] 
 [User Settings]