Is there any sort of known upper limit for the maximum number of tags you can assign to a single tag group? The system integrator onsite is saying that with a 3s poll rate and some 30k+ tags assigned to a single tag group the tag group is being overloaded and causing data loss to the Ignition OPC server. This doesn't seem to make too much sense to me, but they've said that either splitting up the tag groups or changing to subscribed polling seems to alleviate this.
What do you mean the "data loss"? I think you are mentioning the tag history. beside the basic Tag Group setting， you need check the Historical Tag Group setting instead. the rate of Historical Tag Group should larger than the rate of Basic Tag Group.
We've got maybe 250-300k on a single tag group at one site... It would depend on your hardware I'd think. That site mentioned has a split gw architecture each with 24GB+ memory and 12 cores, from memory, dedicated to Ignition
Ignition OPC tags that are in a group together and pointing at a given OPC server will operate together for that OPC connection, either in one subscription, or one read operation per poll.
(You should always be using subscriptions, unless the target OPC server has a problem with that--some do.)
I'm trying to better understand with site what exactly they are meaning when they say "data loss".
As far as subscription mode goes that would be my preference as well, but they are saying the downside of that approach is that the tags remain at their last known state when a device has no power, which is misleading.
That's not how devices nor OPC UA connections to other servers behave. At the very least a non-good quality is applied to the tags once the connection loss is detected.
1 Like
With 300k tags in a single tag group are you using subscribed instead of direct polling?
Keep in mind there's multiple things at play here. There are the drivers that are polling at the tag group's specified poll rate which will happen regardless of if OPC is in polled or subscribed mode. If you're overloading the end device, that could be part of the problem, but you mentioned that the integrator said the problem is fixed by going to subscribed mode. This won't change the way the data is collected, but only how it's reported to the OPC client. I always use subscription mode and have never had a need for polled mode (although I'm sure there are use cases, I would guess the majority of the time subscribed mode works just fine). As @Kevin.Herron pointed out, the data gets marked as bad quality, so while the value may not update, you will know that the data is bad/stale.
Yeah some of the feedback coming back from site doesn't seem to line up. Are there any suggested log files we could pull to try and nail down what is happening?
I would look at the driver diagnostic screens that show the response times of the PLC, or the OPC client diagnostics for the tag groups.
Here's what the device diagnostics will look like (this is an Allen Bradley PLC connection):
This is the OPC diagnostics for the Ignition OPC server:
And this is the diagnostics for a tag group showing status (so you can see if maybe there's a bunch of bad tags that are polling possibly):
They will each show a filtered log view for you to see if there's any errors on each of those pages.
These diagnostics, and the server diagnostics, only apply to IA's native drivers. The OP hasn't indicated whether they are using IA's drivers, or connecting to external OPC servers. (Or both, perhaps.)