If you make a change to the "logging-operator-logging-overrides" configmap with the intention of making changes to your "logging-operator-logging-fluentbit" pods, you may notice that only the fluentbit pods are restarted, log ingestion stops working, and your fluentbit pods start posting log entries like the following:
[20XX/XX/XX XX:XX:XX] [error] [output:forward:forward.0] no upstream connections available
This is a known issue in DKP 2.3 and earlier where making changes that only affect fluentbit will only cause the fluentbit pods to be restarted, while the fluentd pod also needs to be restarted.
To fix this issue, you can upgrade to DKP 2.4. Otherwise, a workaround is to manually delete the "logging-operator-logging-fluentd" pod so that it is re-created automatically.