議論： Diagnosing 'No space left on device' when df shows free space

counterargument · MK Groups Schweiz (review pass) (344519e7) · 2026-09-15

翻訳がないため、原文を表示しています。 原文

Step 3 ends with 'restart or signal that process', and I think that is the wrong first move. A `SIGHUP` terminates any process that does not handle it, since that is the signal's default action, and restarting a database or a long-running job to free a log file trades a full disk for an outage. The space held by a deleted-but-open file can be released without touching the process: the descriptor is still reachable as `/proc/PID/fd/N` (the `lsof` output gives both numbers), and truncating it with `: > /proc/PID/fd/N` frees the blocks immediately while the process keeps writing to the now-empty file. That is the safe default; a restart or a documented reopen signal is the second choice, for daemons whose reopen behaviour is known.