What data is shared in redundant H system and when?

In an H-system CPU with MSTR LED lit is "in control". However, this doesn't mean that the ethernet CP in this CPU's rack is active. It could very well be that the CP from the partner CPU (MSTR LED not lit) is active. Well, at least not in the case of S7-fault tolerant connection. (I don't know if there is any other way to set up a "switched" ethernet connection in H-system)S7-fault tolerant connection only provides a single connection ID.

So, basically your block has to coomunicate only on the basis of connection ID. Hardware address should not (and must not) be a parameter of your block. Is it possible for your driver block?

The question now is, that can you actually configure a S7-fault tolerant connection in Net Pro to a non-siemens device.

>So, basically your block has to coomunicate only on the basis of connection ID.
>Hardware address should not (and must not) be a parameter of your block. Is it possible for your >driver block?
No it is not as I use the standard AG_XXXX calls. Each AG_XXXX call must have an ID and LADDR specific to the CP you wish to use. Are there H-system compatible versions of AG_XXXX functions?

>The question now is, that can you actually configure a S7-fault tolerant connection in Net Pro to a >non-siemens device.
You can't setup a S7-fault tolerant connection to a non-siemens device over TCP (well not to my knowledge). Hence why I am supplying both CP details for each rack, with my block detecting which rack is "in control" and selecting the correct CP paremeters accordingly. This should not be an issue, I have a plan that I am certain will work.

>(The point that I try to make is that two CPUs are not your problem, but two CPs are)I cannot share the block in its entirety due to it being propriety information, but I could give out small sections of it. But I don't think it is relevant. You could just take the case of running a single AG_XXXX call followed by and BLKMOV call to shift the data into a DB and that would demonstrate the problem just as well. Take the following code:-

[code]
AG_LRECV(ID:=RACK0_ID, LADDR:=RACK0_LADDR, RECV := Buf, NDR:=NDR, ERROR:=ERROR, STATUS:=STATUS, LEN:=LEN);
RetVal := BLKMOV(SRCBLK:=Buf, DSTBLK:=DB1);
[/code]
Where RACK0_ID is the ID for the connection for rack 0 CP, RACK0_LADDR is the hardware address for the CP in rack 0 and Buf is an array in the static memory area for the block.
If the code was run in a simplex system it would work fine.
What would happen if this code was run in a h-system dual redundant system?
Remember the ID and LADDR for the CP in rack 1 will be different to the ones for the CP in rack 0.
In the rack 0 CPU the AG_XXXX call would work and the data would be received and the DB would be updated OK.
In the rack 1 CPU the call would fail (wrong ID/LADDR parameters) and Buf would contain wrong data which is now copied to the DB.
So the DB contents of each CPU will be different, and require synchronisation (with hopefully the master CPU values being written to the the reserve CPU).

Dave.

But I see something that I need to point out urgently.

AG_XXXX blocks require a configured connection but do not work with S7 Connection or S7 Fault tolerant connection. (I had earlier mis-interpreted your post and thought that you were not using them, now it's clear)

Unfortunately H-system do not support open communication via Industrial ethernet (TCP, ISO on TCP, UDP etc. in Netpro)

Please use Blocks supported by S7-connections for your protocol driver. (BSEND, BRCV etc.)

Hi Dave,

me again..

You cannot have a simple TCP connection in H-system. It has to be S7 connection, S7 Fault tolerant connection (Iinherently using TCP protocol, but S7 connection in NetPro) or ISO transport connection.

Contrary to what I posted earlier H-system may support AG_XXXX blocks with ISO transport. connection. (not TCP though)

Ok, I believe Rack 0 CPU can be MSTR and communication can happen using Rack 1 CP (and vice versa) in redundant mode. This is true at least in case of S7 fault-tolerant connection and I believe should also be the case when using single sided S7 or ISO transport connection i.e if there is no error.

You can approach the method ofAPRED libraryto communicate to your external device or as described in the manual in section 11.6.1 for single sided S7 connectioni.e it is a better to call one block for one connection and conditionally calling the second block configured for second connection, when there is communication error in the first block. (In my view it is irrelevant which CPU is master or which is reserve) You have to diagnose connections and not CPU state.

Hope it solves some queries.

Hello Smiffy,

I must confess I wasn't able to catch up with all of what had been said in this thread but the question is very good and leads to a good discussion as well.

There are some points I want to share,
1-

smiffy

1) For starters I need to detect which CPU is "in control", and if I am that CPU. As only the CPU in control must run the communications link, otherwise both CPU's/CP's will be asking for data. I have a solution for this using the RDSYSST SFC and SSL-ID of 16#0174 (Read LED status). My block will obviously need the CP connection parameters (NETPRO ID and hadware address) for each CP. Then it would use the relevant CP connection parameters based on which CPU is currently "in control".

An easier way is to use the pre made block to distinguish between master and standby CPUs.

2-

smiffy

Can anyone confirm that the master DB will overwrite the standby DB?

I've searched the manual quickly to find any piece of information saying that master CPU transfers values of data blocks in standby CPU.

This is what typically happens in software redundancy but it's totally different since only one of the CPUs execute the application program and the other just receives data and use the last received data at switching over (Don't ever ask about data transfer time).

H-System technique (Active redundancy) assumes that both CPUs execute the same and typical piece of application program.
Execution is synchronized using event driven synchronization and this means that probability of time shift in SW execution between both CPUs is very small.
So when switching over takes place there's no probability that the values in data blocks are little old or little different (They should be the same of the master data blocks).

There's a phase called UPDATE (extracted from Link UP & Update) and in this phase the master CPU transfers contents of data blocks to the standby CPU (Starting CPU).
If data blocks are exchanged at each cycle then what's the use of update phase?!

Now, you want to jump over some instructions in standby CPUand change values in a data block inmaster CPUs. No one can tell you the results. You have to test it yourself.

3-

smiffy

But I would expect that the H-system synchronisation scheme would keep the 2 DB's in sync even if only one CPU updated the DB. Surely the dual-redudant scheme is not solely relying on the assumption that both CPU will be running the same code in the same way, so therefore the memory areas (and DB's) cannot get out of sync? If so that is a very poor way of doing things.

As I told you the technique assumes that there's no probability to have different values in data blocks if both CPUs are reading the same inputs and taking the same decisions at the same time.

My explanation would be that it will take much time to exchange values of data blocks at each cycle(S7-400H project most likely means many data blocks with big volumes of data).

Well, let's hope I'm wrong but being wrong or right will never change the fact that H-system is very powerful and has proven itself in the market over the whole world for a long long time since S5-H.

4-
I hope you'll be able to do what you want to do but there's one solution which you may put in your pocket.
Siemens says you can use PC with CP1613 as a gateway. This gateway allows you to link any kinds of single-channel system (e.g. TCP/IP with a manufacturer-specific protocol).
Look at the attached PDF.

Best regards.

H-H