The default maxSurge value for k8s deployment is 25%, and when replicas is 1, maxSurge is 1 after rounding up. This results in the existence of 2 identical edgex components at the same time after a quick create + update operation, and edgex components use a registration mechanism. If a component preregisters and is terminated by workload, this causes the metadata component to scramble. This chaos manifests itself in the form of other components crashing all the time.
I updated auto-collector to directly inject the default maxSurge value of 0 into the config file and the issue was resolved.
What type of PR is this?
/kind bug /sig iot
What this PR does / why we need it:
The default maxSurge value for k8s deployment is 25%, and when replicas is 1, maxSurge is 1 after rounding up. This results in the existence of 2 identical edgex components at the same time after a quick create + update operation, and edgex components use a registration mechanism. If a component preregisters and is terminated by workload, this causes the metadata component to scramble. This chaos manifests itself in the form of other components crashing all the time.
I updated
auto-collector
to directly inject the default maxSurge value of 0 into the config file and the issue was resolved.Which issue(s) this PR fixes:
Fixes #2028