Ignition 8.1.49, redundant gateway pair.
We have a redundant PLC pair (Hot Standby). The CPU has a floating IP that
follows the primary, so anything talking straight to the CPU rides through a
switchover without noticing. The OPC UA server, though, is a module sitting
in each rack, and those modules do not share that floating IP β each has its
own fixed IP and its own endpoint. Ignition connects to the modules.
Reads work against either server. Writes only work against the one in the
rack holding the primary CPU. So the gateway has to keep its tags pointed at
the right server, and which one is right changes on every switchover.
Configuring the second endpoint as the failover of a single connection
covers the loss of a module, but not the case where both servers are healthy
and only the primary role moved.
What we ended up doing:
- Two independent OPC UA connections, one per server, no failover between
them. - Tags reading the Hot Standby status from each server.
- An expression tag that resolves which connection should be in use.
- A parameter in every UDT holding the OPC server name, with every OPC
member tag binding its OPC Server property to that parameter. - A reference tag in each UDT that follows the expression tag and writes the
resolved name into its own instance parameter on value change.
It works, but it means that on every switchover around 1,200 UDT instances
rewrite a configuration parameter at runtime, reconfiguring something like
27,000 tags. The writes are asynchronous with no confirmation and no retry.
It ran clean for about a year. Recently we had two events where part of the
tag tree stayed on the standby server even though the expression tag had
already resolved correctly β values visible, commands dead. Restarting the
tag and both connections brought it back.
What we would really want is to express this at the connection level. The
server module reports a ServiceLevel of 255 when it is the primary and lower
when it is the standby, so something along the lines of "fail over when
ServiceLevel < 255" would let the driver pick the endpoint and leave the
tags alone. ServiceLevel is just the angle we thought of β any approach that
avoids changing tag configuration at runtime would do.
Has any of you run into a similar problem? Curious how others handle
redundant PLC pairs where the OPC UA server does not follow the primary.
