| Author |
Message |
Rob M
Guest
|
 Posted:
Sun Nov 13, 2005 1:50 am Post subject:
Enterprise Edition Failover |
|
|
I have a test setup with a Domain Controller, 2 Enterprise Edition servers
(LCS 2005 with SP1) and 2 clients with Communicator, all running on Windows
Server 2003. The 2 LCS servers are using Network Load Balancing.
Stopping one of the LCS servers (eg by using Net Stop RTCSRV) should cause
clients on that server to be rehomed on the other pool server, however at the
moment this is not happening. Also, clients that were logged on to the server
that is stopped can not sign in. However, shutting this server down
completely allows clients to log on to the pool using the remaining server.
Does anyone know how the inter-server communication happens? There doesn't
seem to be any documentation on this. |
|
| Back to top |
|
 |
Peter Haugaard Larsen
Guest
|
| Posted:
Mon Nov 14, 2005 1:50 am Post subject:
Re: Enterprise Edition Failover |
|
|
Hi Rob
When you are using NLB to load balance it dosn't look at the services or
ports you are loadbalancing to see if they are online.
It only removes a server from the NLB cluster if the server stop sending a
heartbeat signal for 5 seconds (default time).
If you want to test NLB failover it is recommended to use the wlbs.exe
commandline utill.
Microsoft recommend you to use hardware loadbalancing and to test on the
service ports, 5060/TCP for non TLS LCS, 5061/TCP for TLS LCS and 135/TCP
for managing LCS.
It's all explained on page 45 and forward in the LCS 2005 Planing guide.
http://www.microsoft.com/downloads/details.aspx?FamilyId=F7BC430F-3CAC-4DBD-8EC3-B93186F343FA&displaylang=en
Regards,
Peter Haugaard Larsen
"Rob M" <RobM@discussions.microsoft.com> wrote in message
news:D94B06E6-ADDB-4A3F-BE05-CB0666ADEF36@microsoft.com...
| Quote: | I have a test setup with a Domain Controller, 2 Enterprise Edition servers
(LCS 2005 with SP1) and 2 clients with Communicator, all running on
Windows
Server 2003. The 2 LCS servers are using Network Load Balancing.
Stopping one of the LCS servers (eg by using Net Stop RTCSRV) should cause
clients on that server to be rehomed on the other pool server, however at
the
moment this is not happening. Also, clients that were logged on to the
server
that is stopped can not sign in. However, shutting this server down
completely allows clients to log on to the pool using the remaining
server.
Does anyone know how the inter-server communication happens? There doesn't
seem to be any documentation on this.
|
|
|
| Back to top |
|
|
Rob M
Guest
|
| Posted:
Mon Nov 14, 2005 5:50 pm Post subject:
Re: Enterprise Edition Failover |
|
|
Thanks for the comments. I have got the planning document, but hadn't had
time to read it fully.
The behaviour I am seeing is that when stopping the RTCSRV service on one of
the two LCS servers, I would expect that a client connected to that server
should automatically redirect to the other server. This is happening, but not
reliably.
I suspect that in my impatience to query the RTC database in SQL Server, I
am not allowing enough time for the absence of the heartbeat to be picked up
- in other words as long as I allow at least 10-20 seconds the failover
should work correctly under this setup (all of the servers are running under
Virtual PC).
"Peter Haugaard Larsen" wrote:
| Quote: | Hi Rob
When you are using NLB to load balance it dosn't look at the services or
ports you are loadbalancing to see if they are online.
It only removes a server from the NLB cluster if the server stop sending a
heartbeat signal for 5 seconds (default time).
If you want to test NLB failover it is recommended to use the wlbs.exe
commandline utill.
Microsoft recommend you to use hardware loadbalancing and to test on the
service ports, 5060/TCP for non TLS LCS, 5061/TCP for TLS LCS and 135/TCP
for managing LCS.
It's all explained on page 45 and forward in the LCS 2005 Planing guide.
http://www.microsoft.com/downloads/details.aspx?FamilyId=F7BC430F-3CAC-4DBD-8EC3-B93186F343FA&displaylang=en
Regards,
Peter Haugaard Larsen
"Rob M" <RobM@discussions.microsoft.com> wrote in message
news:D94B06E6-ADDB-4A3F-BE05-CB0666ADEF36@microsoft.com...
I have a test setup with a Domain Controller, 2 Enterprise Edition servers
(LCS 2005 with SP1) and 2 clients with Communicator, all running on
Windows
Server 2003. The 2 LCS servers are using Network Load Balancing.
Stopping one of the LCS servers (eg by using Net Stop RTCSRV) should cause
clients on that server to be rehomed on the other pool server, however at
the
moment this is not happening. Also, clients that were logged on to the
server
that is stopped can not sign in. However, shutting this server down
completely allows clients to log on to the pool using the remaining
server.
Does anyone know how the inter-server communication happens? There doesn't
seem to be any documentation on this.
|
|
|
| Back to top |
|
|
Peter Haugaard Larsen
Guest
|
| Posted:
Mon Nov 14, 2005 5:50 pm Post subject:
Re: Enterprise Edition Failover |
|
|
Hi Rob
You'll need to kill the server it self before it stops sending heartbeats.
There after you can do a manual logout/login, for the client to automatic
failover you'll have to wait about 5 minuts.
Regards,
Peter Haugaard Larsen
"Rob M" <RobM@discussions.microsoft.com> wrote in message
news:F2975ED4-6907-46C2-8BED-73CD7F12BDEF@microsoft.com...
| Quote: | Thanks for the comments. I have got the planning document, but hadn't had
time to read it fully.
The behaviour I am seeing is that when stopping the RTCSRV service on one
of
the two LCS servers, I would expect that a client connected to that server
should automatically redirect to the other server. This is happening, but
not
reliably.
I suspect that in my impatience to query the RTC database in SQL Server, I
am not allowing enough time for the absence of the heartbeat to be picked
up
- in other words as long as I allow at least 10-20 seconds the failover
should work correctly under this setup (all of the servers are running
under
Virtual PC).
"Peter Haugaard Larsen" wrote:
Hi Rob
When you are using NLB to load balance it dosn't look at the services or
ports you are loadbalancing to see if they are online.
It only removes a server from the NLB cluster if the server stop sending
a
heartbeat signal for 5 seconds (default time).
If you want to test NLB failover it is recommended to use the wlbs.exe
commandline utill.
Microsoft recommend you to use hardware loadbalancing and to test on the
service ports, 5060/TCP for non TLS LCS, 5061/TCP for TLS LCS and 135/TCP
for managing LCS.
It's all explained on page 45 and forward in the LCS 2005 Planing guide.
http://www.microsoft.com/downloads/details.aspx?FamilyId=F7BC430F-3CAC-4DBD-8EC3-B93186F343FA&displaylang=en
Regards,
Peter Haugaard Larsen
"Rob M" <RobM@discussions.microsoft.com> wrote in message
news:D94B06E6-ADDB-4A3F-BE05-CB0666ADEF36@microsoft.com...
I have a test setup with a Domain Controller, 2 Enterprise Edition
servers
(LCS 2005 with SP1) and 2 clients with Communicator, all running on
Windows
Server 2003. The 2 LCS servers are using Network Load Balancing.
Stopping one of the LCS servers (eg by using Net Stop RTCSRV) should
cause
clients on that server to be rehomed on the other pool server, however
at
the
moment this is not happening. Also, clients that were logged on to the
server
that is stopped can not sign in. However, shutting this server down
completely allows clients to log on to the pool using the remaining
server.
Does anyone know how the inter-server communication happens? There
doesn't
seem to be any documentation on this.
|
|
|
| Back to top |
|
|
|
|
|
|
FAQ
Memberlist
Register
Profile
Log in to check your private messages
Log in