Redundent WinCC OA Servers

[ilyas_ahmed]

Hi Experts,

Is it possible to use redundant hot standby configuration with both servers at different geographical locations e.g. at approx. 4km apart linked with each other on a GHz fiber optic network?
Will it effect the time system needs to switch to standby in case primary server fails.

Thanks!

[leoknipp]

It is possible to use a redundant WinCC OA system with servers in different geographical locations.
The requirement is that you have a stable and fast network connection for the TCP/IP-messages.

Best Regards
Leopold Knipp
Senior Support Specialist

[ilyas_ahmed]

Hi,

Thanks for that, my next question is about the "Value archiving" capacity of each WinCC OA servers, in case I don't use oracle clusters. In other words whats the maximum number of tags / size of local database we can go without effecting the performance of the servers. Lets assume 1000 tags for approx 5 years. Do we need to consider an Oracle Cluster.
Still considering that the redundant servers are at different locations.

Thanks

[leoknipp]

The amount of historical data saved on the disc does not have an impact on the performance during runtime.
When reading historical data it can make a small difference how data is organized, depending on the configuration of the archives.

When saving historical values you have to take into account how many bytes are needed to store a value/alarm.
Information you find in the following FAQ
https://portal.etm.at/index.php?view=it ... &Itemid=54

Best Regards
Leopold Knipp
Senior Support Specialist

[NikolayLevchenko]

Having redundant servers in different locations is not a problem in my experience.
One thing I learned from this that you might need to consider is the following.

If you have Server A in location A and Server B in location B and then multiple clients in locations C, D, E, etc. you mmight have the following scenario:
1. Server A ACTIVE, all clients use it to write commands
2. Network between location C and A goes down.
3. Since the network between A and B is still OK, server A stays ACTIVE
4. Your clients see that they lost connection to server A and switch to server B.
5. Now your clients are useless since any commands they send go to server A and server B blocks them since it's PASSIVE.

So I solved it by adding to a weight calculation on the servers the connection status to UI's. Now when server A looses connection to UI but server B still have them, server B becoms ACTIVE.

See this thread: https://portal.etm.at/index.php?option= ... id=16#1915

[uxout]

Hello,
I'm having another question linked to this one.
In the same case, if no client is connected to server A, so the server A is active.
My clients couldn't connect anymore, because only server B (passive) is available.
How can you resolve this issue?

[NikolayLevchenko]

I'm not sure I understand your situation. Can you provide more details?

[uxout]

Following this schema : Click here
The switch from Server A to clients is down but there is no way to know as no clients is connected to the servers.
Therefore, Server A is still active and Server B is passive.

The client will still try to connect to Server A and fails.
Afterward the client will try to connect to Server B, succeed, but receive an error as its a passive server.

What could be the solution to this case?

[leoknipp]

If only the UI connection is used to calculate the server state and no UI is running the server cannot be switched.
To enable an automatic switch to the server which has the better state you have to add additional parameters to the error state calculation, e.g. the connection to PLCs, ...

Best Regards
Leopold Knipp
Senior Support Specialist

[NikolayLevchenko]

Normally your UI should always connect to BOTH redundant hosts (check out config files entry "connectToRedundantHosts").
If connection to Server A fails, the connection to server B is still working. In this case, the client becomes a bit useless because any commands you send from it will go to PASSIVE server and get discarded there.

So what you can do is configure error weight on pre-defined UI connections. This way (in the drawing you provided) server A will have non-zero error weight but Server B will have zero error weight. This will trigger a switch-over and Server B will become ACTIVE. Voila!