https://bugzilla.kernel.org/show_bug.cgi?id=197335 Bug ID: 197335 Summary: Inode exhaustion should be logged to dmesg Product: File System Version: 2.5 Kernel Version: 4.14.0-rc1 Hardware: All OS: Linux Tree: Mainline Status: NEW Severity: enhancement Priority: P1 Component: ext4 Assignee: fs_ext4@xxxxxxxxxxxxxxxxxxxx Reporter: teamathena.nitc@xxxxxxxxx Regression: No Problem: If the system runs out of inodes, the error messages show "No space on disk". This is misleading to the user as the disk may still have space. Logging to dmesg can help the user correctly identify the problem. Bug #1665833 from https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1665833 My server just became unusable, because it ran out of inodes (something that shouldn't happen on a relatively normal installation). The root-cause was a misbehaving CRM program creating way too many session files and never deleting them. However, while debugging, it would have been really really helpful if there were something in the logfiles. But it just failed silently: the lack of any error message makes diagnosing this error condition 100x harder than it should be. We get messages about being out of disk-space, but that's not helpful, especially when "df -h" shows it not to be true. So, the bug report/feature request is this: when the kernel cannot create a new file due to a shortage of inodes, it should emit a message to that effect, at least in dmesg (and ideally also to syslog). [For anyone else who chances upon this bug report, you can tell if you have run out of inodes with "df -i", and you can find the offending directories with: "find / -xdev -size +100k -type d" ] -- You are receiving this mail because: You are watching the assignee of the bug.
