rsync --partial vs --inplace: choose by recovery and reader behavior
Rsync partial vs inplace: inspect the effective state, interpret expected and abnormal results, make one bounded change, verify, and roll back safely.
Use SFTP batch mode with explicit failure and atomic-file behavior
SFTP batch mode: inspect the effective state, interpret expected and abnormal results, make one bounded change, verify, and roll back safely.
Use SSH BatchMode so automation fails instead of waiting for input
SSH BatchMode: inspect the effective state, interpret expected and abnormal results, make one bounded change, verify, and roll back safely.
Use SSH Match exec without turning config into an opaque program
SSH Match exec: inspect the effective state, interpret expected and abnormal results, make one bounded change, verify, and roll back safely.
Use SSH ControlPersist in automation without stale shared masters
SSH ControlPersist automation: inspect the effective state, interpret expected and abnormal results, make one bounded change, verify, and roll back safely.
Use ExitOnForwardFailure so SSH tunnels fail loudly
ExitOnForwardFailure: inspect the effective state, interpret expected and abnormal results, make one bounded change, verify, and roll back safely.
Use ssh-keyscan for collection, never as automatic verification
Ssh-keyscan safe use: inspect the effective state, interpret expected and abnormal results, make one bounded change, verify, and roll back safely.
Automate SSH ProxyJump without placing private keys on the bastion
SSH ProxyJump automation: inspect the effective state, interpret expected and abnormal results, make one bounded change, verify, and roll back safely.
ServerAliveInterval vs TCPKeepAlive for SSH connection liveness
ServerAliveInterval vs TCPKeepAlive: inspect the effective state, interpret expected and abnormal results, make one bounded change, verify, and roll back safely.
Use StrictHostKeyChecking accept-new with a clear trust model
StrictHostKeyChecking accept-new: inspect the effective state, interpret expected and abnormal results, make one bounded change, verify, and roll back safely.