6

Consider the following scenario:

  • I use Linux device mapper to create a snapshot of an ext4 file system.
  • The snapshot is mounted as read-only; the source volume is mounted as read-write.
  • I read the snapshot, and simultaneously write (too much) to the source volume. Eventually, the copy-on-write table fills up.

Now exactly what happens in practice from the user's perspective? What kind of messages should I expect to see in dmesg? How do the applications that read the snapshot behave? Has someone actually tried this to see what would happen?

Jukka Suomela
  • 159
  • 6
  • This looks interesting, but I don't understand why you ask that instead of trying it yourself - since you know exactly how to trigger such situation. – rozcietrzewiacz Aug 17 '11 at 22:00
  • @rozcietrzewiacz: Not that easy right now, unfortunately... Besides, I suspect that the answer is non-unique and may depend on the exact circumstances; hence I would be happy to hear of any experiences that other people have had with this kind of situations. – Jukka Suomela Aug 17 '11 at 22:22
  • 3
    Did you read this Q&A: [LVM is reporting I/O errors, but the disk reports no problems. Argh](http://serverfault.com/questions/189803/lvm-is-reporting-i-o-errors-but-the-disk-reports-no-problems-argh) ? – Mat Aug 18 '11 at 04:58
  • @Mat: Thanks, this is certainly relevant; perhaps you could post it as an answer (link + a brief summary)? – Jukka Suomela Aug 18 '11 at 09:10

1 Answers1

3

When the COW fills up, you start getting I/O errors on write operations.

LVM2 allows you to check the size and usage of the COW, and resize it if necessary.

Alan Porter
  • 226
  • 1
  • 4