TopolearnFull disks and memory pressure are among the most frequent causes of outages.
df -h shows each mounted filesystem's size, used and available space in human-readable units. df -h /var shows just the filesystem that holds /var. A filesystem can also run out of inodes (one per file) while bytes are still free, typically with millions of tiny files; check with df -i.
du measures what directories contain: du -sh dir gives one total, and du -h --max-depth=1 /var | sort -h lists subdirectories by size so you can drill down. ncdu is an interactive alternative. If df reports a full disk but du cannot find the space, a process may still hold a deleted file open (often a rotated log); lsof +L1 lists such files, and restarting the process frees the space.
free -h shows memory. Linux deliberately uses spare RAM as page cache for files, so "free" is often small on a healthy machine. The number that matters is available: memory that can be handed to programs, including reclaimable cache. Swap is disk space used as overflow; some swap use is fine, but constant swapping in and out makes everything slow.
On servers, keep an eye on /var (logs, databases, Docker images in /var/lib/docker) and set up log rotation and monitoring before the disk fills.
Example
df -h # free space per mounted filesystem
df -h /var # only the filesystem holding /var
df -i / # inode usage: "No space left" with free bytes?
du -sh ~/Downloads # total size of one directory
sudo du -h --max-depth=1 /var 2>/dev/null | sort -h | tail -n 5 # biggest subdirs
sudo lsof +L1 # deleted files still held open
free -h
# total used free shared buff/cache available
# Mem: 15Gi 4.1Gi 1.0Gi 0.5Gi 10Gi 10Gi
# Swap: 2.0Gi 0B 2.0GiRemember
- df shows free space per filesystem, du shows what uses it
- df -i for inodes; lsof +L1 for deleted-but-open files
- In free, read 'available', not 'free'
Watch out for
- Panicking because
freeshows little free memory, when most of it is reclaimable page cache.