Conditions That Abort Ansible Rolling Update

Quick Answer

any_errors_fatal: yes causing an immediate abort is correct because that setting changes Ansible's default failure handling at the most fundamental level: normally Ansible tries to continue running a play on the hosts that have not failed, but any_errors_fatal overrides that and says a single task failure on any single host is enough to stop the entire play for every host, immediately, with no regard for batch size or percentage thresholds. That makes it meaningfully different from max_fail_percentage, which the explanation also describes as an abort trigger - max_fail_percentage only aborts once the proportion of failed hosts in a batch crosses a set threshold, so it tolerates some failures depending on how that percentage is configured, whereas any_errors_fatal tolerates none, at any point in the run. Both are legitimate ways to make a rolling update stop early, but they operate on different logic: one is threshold-based and batch-aware, the other is absolute and immediate. When a question asks about conditions that abort a rolling update, look for whether the described trigger is proportional, a percentage of a batch failing, or absolute, any failure at all regardless of batch, since those map directly to max_fail_percentage and any_errors_fatal respectively.