Bug: swss/orchagent "config reload" failed on armhf #24766
Description
Is it platform specific
marvell
Importance or Severity
Critical
Description of the bug
Platform: marvell-prestera ARCH: armhf Board Nokia-7215
Branch - TRIXIE (not branch 202505)
The board has week-CPU.
"config reload" failed under High CPU-loading (for example - PTF test_po_cleanup.py)
case CPU-loading time-to-ready Success/Fail(%)
IDLE 55% 350 sec 100%-success
2 "yes" processes full-100% 870 sec 5%-success
2 "yes" + Workaround full-100% 550 sec 100%-success
Deep-trial-investigation:
The "config reload" makes restart for the sonic.target superset service.
The superset start chain/dependency sequence is asynchronous.
It works perfect on Fast-CPUs, but has "bad/wrong async timing"
between swss/syncd and OTHER services started in the set
on week CPU armhf Nokia-7215.
Under debugging, the workaround has been found:
- Stop swss and syncd services at once after "restart sonic.target"
- restart swss
Steps to Reproduce
1). Run PTF test_po_cleanup.py on Nokia7215(armhf) board with TRIXIE
2). manually step-by-step:
taskset 0x1 yes > /dev/null 2>&1 &
taskset 0x2 yes > /dev/null 2>&1 &
config reload -y
wait 15 min
docker exec swss supervisorctl status orchagent
show interfaces status
Actual Behavior and Expected Behavior
Relevant log output
Output of show version, show techsupport
Attach files (if any)
No response