I have a pretty basic question, perhaps I didn't know the right keywords as I couldn't find a previous answer. I use Python scripts control and gather information for a smarthome environment. I mostly use text files to store and update information within and between the scripts. However, I frequently run into this one issue whenever the server crashes or loses power: The file contents tend to corrupt or vanish while the crash happens.

To write file content, I usually use a structure like this:

try:
    with open(savefile, "r") as file:
        lines = file.readlines()
except:
    lines = []
    pass
lines.append(str(time.time()) + ";" + str(value) + "\n")
if len(lines) > MAX_READINGS:
    lines = lines[-MAX_READINGS:]
with open(savefile, "w") as file:
    file.writelines(lines)

In case of partial corruption such as blank lines between the data points, I often use a line-by-line loop that only qualifies lines with the correct structure (such as a timestamp in the beginning in the example-like cases). However, sometimes a file gets corrupted to the point it only contains spaces or is empty, getting useless for the scrips depending on the data.

The filesystem's integrity remains intact in crashes, so it's probably not a lower level problem. But what's the suggested workaround to minimize the corruption risk?

Should I use the "a" mode to append a new line and have another way to deal with the file lengths (the MAX_READINGS), or should I make a temporary copy which I'd then use to overwrite the original after the writing is done? Or might there be an external library providing the right functionality?