This function receives a range vector and a scalar time value as arguments. It extrapolates the value of each matched time series from the query evaluation time to the specified number of seconds in the future, given the trend in the data from the range vector. It uses linear regression to achieve such a prediction, which means there is no complex algorithmic forecasting happening in the background. It should only be used with gauges.

We'll apply the following expression, which employs predict_linear using a range of one hour of data, and extrapolate the sample value four hours in the future (60 (seconds) * 60 (minutes) * 4):

predict_linear(node_filesystem_free_bytes{mountpoint="/data"}[1h], 60 * 60 * 4)

{device="/dev/sda1", endpoint="node-exporter",fstype="ext4",instance="10.0.2.15:9100", job="node-exporter-service",mountpoint="/data", namespace="monitoring", pod="node-exporter-r88r6", service="node-exporter-service"} 15578514805.533087