Running into an issue where all of my SQL Bridge threads clock up for readBlocking tasks from transactions. This ends up causing a domino effect where all of my transactions hold up until a timeout occurs.
I’ve ran into this on two separate gateways with different devices now, and I’m struggling to find a good solution.
In this particular instance, PLC memory and CPU usage are low, the numbers showing on the ignition device page seem good, and wireshark didn’t seem to show any obvious signs of issues. The system has 57 Transactions all set to Read mode, which ends up only subscribing to 58 tags total.
I’ve tried increasing concurrent requests to the PLC, opening more threads for SQL Bridge, lowering the OPC read timeout. All of which don’t seem to fix any issues.
I’ve got a ticket open with IA, but I haven’t heard anything back just yet, and wondered if anyone else in the community has an idea of what to look for.
Ignition-JR1QZRRQ21823_thread_dump20200312-120026.txt (195.8 KB)
All of the items in OPC Read mode (other than the subscriptions for triggers) are invisible to the device loading measurements on the status pages. They can only show the bandwidth and performance of subscription items. Reads that don't use a subscription (anywhere else in Ignition, too) cannot be tracked, nor the bandwidth for writes.
You may simply be overloading the PLC connection. Consider putting all of your transactions temporarily in subscribed mode just to see the resulting statistics.
I may do that just to paint a better picture of the load on the system, but we see these threads getting locked up even without the transactions running. So only the trigger tags and our heartbeat tag should be getting polled.
After putting all transactions to subscribe, this is what our device page looks. The average read time is fairly low even with all the tags polling at the same rate. Device load went up a bit, but was still isn’t too high.
Hmmm. Single-digit millisecond response times, or nearly so. Nothing wrong with your connection. Might be time to ping your support ticket.
Just to keep the thread updated in case anyone else ends up having the issue, we exported our transactions and pulled them into a 7.9.13 gateway running all default configurations for threads and OPC settings, and we are not seeing these threads being stopped anymore.
Planning to continue monitoring and follow up with support this morning once they open.
