We’ve run into an issue where a Haystack connection is substantially raising the CPU usage on a Jace. The CPU is pegged at/near 100% whenever the Haystack connection is established. We’ve tried different combinations of Niagara 4.4, 4.7 and 4.9 with nHaystack 2.1.2 and 3.0.1 all with the same results (there is slightly less CPU usage on 4.4/2.1.2, but still greater than 90%)
Has anyone else run into a similar issue? Or have any idea of the cause?
Alper ÜzmezlerThu 10 Dec 2020
Keith,
Call me since you have an environment that is in place. Let's enable spy log for these and check errors. If we can't see it, we might need to place log outs for the nhaystack. I will try to help. Also can you create a ticket on github as well.
I have ran into high CPU Usage as well on JACEs with 2 or 3.0.1 with N4.4, 4.7, and 4.9. SoftJaces/Supervisors seem to have enough horsepower to deal with nHaystack queries from SkySpark, but on JACEs I usually resort to a process other than haystackReadAll(...) to build (either exporting portions of the system to csv, or building manually on small sites).
From what I could remember, it seemed like every time I initiated a query against the nHaystack driver, it caused a new login session with the JACE (something in how Tridium handles login sessions). Supposedly using an HTTP Basic Auth Scheme helps with this, but I didn't have much luck.
Would be curious to see the results of your guys investigation, and/or if you want to throw up your troubleshooting steps/what you modify for spy logging and maybe compare notes!
Keith Bishoρ Thu 10 Dec 2020
We’ve run into an issue where a Haystack connection is substantially raising the CPU usage on a Jace. The CPU is pegged at/near 100% whenever the Haystack connection is established. We’ve tried different combinations of Niagara 4.4, 4.7 and 4.9 with nHaystack 2.1.2 and 3.0.1 all with the same results (there is slightly less CPU usage on 4.4/2.1.2, but still greater than 90%)
(These images are with 4.9 and 3.0.1)
Has anyone else run into a similar issue? Or have any idea of the cause?
Alper Üzmezler Thu 10 Dec 2020
Keith,
Call me since you have an environment that is in place. Let's enable spy log for these and check errors. If we can't see it, we might need to place log outs for the nhaystack. I will try to help. Also can you create a ticket on github as well.
Let's track it from here.
https://github.com/ci-richard-mcelhinney/nhaystack/issues
Mike Melillo Wed 16 Dec 2020
I have ran into high CPU Usage as well on JACEs with 2 or 3.0.1 with N4.4, 4.7, and 4.9. SoftJaces/Supervisors seem to have enough horsepower to deal with nHaystack queries from SkySpark, but on JACEs I usually resort to a process other than haystackReadAll(...) to build (either exporting portions of the system to csv, or building manually on small sites).
From what I could remember, it seemed like every time I initiated a query against the nHaystack driver, it caused a new login session with the JACE (something in how Tridium handles login sessions). Supposedly using an HTTP Basic Auth Scheme helps with this, but I didn't have much luck.
Would be curious to see the results of your guys investigation, and/or if you want to throw up your troubleshooting steps/what you modify for spy logging and maybe compare notes!