Hello,
I have a PFC200 controlling a smallish new plant, but the linux load average is very high (over 5 sometimes, which then can lead to a shutdown of the codesyscontrol runtime).
I have a separate PLC setup for testing, as the plant is nearly up and running. The test setup is much smaller, but can produce the same/similar problems.
It is possible that the PLC is simply overworked, but I have noticed that the load average seems to be random:
after a software download, the load average will often rise to 4 or 5, and not come back down.
A reboot MAY fix it, getting the load back to around 1. But sometimes the load will come to 4..5 even after a reboot.
At first I suspected the PFC200bus (local IO-bus), so I disabled it from the test system, leaving the CAN-bus and the EtherCAT active.
After a number of reboots with no other changes to the system whatsoever, I have seen the following load averages after 10 minutes of operation (this is the smaller test setup, described below):
1min/5min average from 'top' command:
1,30/2,19 after software download
1,15/1,15 after power cycle 1. PLCLOAD in codesys: 35%
1,66/1,82 after power cycle 2. PLCLOAD in codesys: 37%
5,92/4,68 after power cycle 3. PLCLOAD in codesys: 38% <--- No changes to setup, only the power cycle
The production system consists of the following:
- 18 CANopen nodes, (only one PDO RX/TX each, 100ms cycle)
- 3 EtherCAT nodes (wago 750-354) with a total of about 65 IO-cards, 10ms cycle
- PFC200Bus (local IO-bus): 7 Digital, 4 Analogue cards, 10ms cycle
- Symbol configuration / OPC UA server is actively used for external logging of all analogue values (about 5s update rate, communication is not encrypted)
- WebVisu used for operation of system, sometimes several users.
My local test setup consists of:
- 1 CANopen node
- 1 EtherCAT node
- PFC200Bus: 3 Analogue input cards
- OPC UA client on my PC
- WebVisu in browser on my PC, 1 client.
I suppose any load average over 1 indicates that too much is going on, but would anyone have an idea what could cause this change in load between reboots, when nothing else seems to change?
Bonus info:
The plc on site is running 3.5.17.10 firmware 4.2.0.0. There seems to be a bug in the CAN driver, which fills the /var/log partition within the first minute or so, so no info from the crashes available.
The test plc is upgraded to 3.5.17.20, firmware 4.3.0.0. This has fixed the CAN issue, so logs are available, but no crashes have been seen so far. The load issues seems to be the same, though.
On my test setup, I have tried a lot of different configurations. Slowing down the EtherCAT task helps, but does not remove the problem. Disabling the EtherCAT does seem to fix it. At least no problems seen with 10 minute load average after 6 reboots.
Have you received any feedback or found any solutions to the load problem. We are running into similar issues, even without codesys running the load avarage is above 1 meaning that it is to busy already without plc code running.
Hope you might have some more insight.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hello,
I have a PFC200 controlling a smallish new plant, but the linux load average is very high (over 5 sometimes, which then can lead to a shutdown of the codesyscontrol runtime).
I have a separate PLC setup for testing, as the plant is nearly up and running. The test setup is much smaller, but can produce the same/similar problems.
It is possible that the PLC is simply overworked, but I have noticed that the load average seems to be random:
after a software download, the load average will often rise to 4 or 5, and not come back down.
A reboot MAY fix it, getting the load back to around 1. But sometimes the load will come to 4..5 even after a reboot.
At first I suspected the PFC200bus (local IO-bus), so I disabled it from the test system, leaving the CAN-bus and the EtherCAT active.
After a number of reboots with no other changes to the system whatsoever, I have seen the following load averages after 10 minutes of operation (this is the smaller test setup, described below):
1min/5min average from 'top' command:
1,30/2,19 after software download
1,15/1,15 after power cycle 1. PLCLOAD in codesys: 35%
1,66/1,82 after power cycle 2. PLCLOAD in codesys: 37%
5,92/4,68 after power cycle 3. PLCLOAD in codesys: 38% <--- No changes to setup, only the power cycle
The production system consists of the following:
- 18 CANopen nodes, (only one PDO RX/TX each, 100ms cycle)
- 3 EtherCAT nodes (wago 750-354) with a total of about 65 IO-cards, 10ms cycle
- PFC200Bus (local IO-bus): 7 Digital, 4 Analogue cards, 10ms cycle
- Symbol configuration / OPC UA server is actively used for external logging of all analogue values (about 5s update rate, communication is not encrypted)
- WebVisu used for operation of system, sometimes several users.
My local test setup consists of:
- 1 CANopen node
- 1 EtherCAT node
- PFC200Bus: 3 Analogue input cards
- OPC UA client on my PC
- WebVisu in browser on my PC, 1 client.
I suppose any load average over 1 indicates that too much is going on, but would anyone have an idea what could cause this change in load between reboots, when nothing else seems to change?
Bonus info:
The plc on site is running 3.5.17.10 firmware 4.2.0.0. There seems to be a bug in the CAN driver, which fills the /var/log partition within the first minute or so, so no info from the crashes available.
The test plc is upgraded to 3.5.17.20, firmware 4.3.0.0. This has fixed the CAN issue, so logs are available, but no crashes have been seen so far. The load issues seems to be the same, though.
On my test setup, I have tried a lot of different configurations. Slowing down the EtherCAT task helps, but does not remove the problem. Disabling the EtherCAT does seem to fix it. At least no problems seen with 10 minute load average after 6 reboots.
Thank you for any insights or suggestions
-Bjarne
Have you received any feedback or found any solutions to the load problem. We are running into similar issues, even without codesys running the load avarage is above 1 meaning that it is to busy already without plc code running.
Hope you might have some more insight.