If the migration was interrupted then it should be possible to just
restart it and it will pick up from where it left off. (It may be
required to run the ipfs-4-to-5
binary directly.)
If the migration was not interrupted but fails for some other reason it will normally attempt to revert back to the original format. If this happens correct the error and try again. If you wish to suppress the automatic reverting you can use '-no-revert' option which will leave the repo in a inconsistent but recoverable state to make it easier to fix the error and then continue by restarting the recovery.
If the migration was interrupted (or aborted due to an error
-no-revert
is used) and restarting it does not work then it should
be possible to get the repo in a state where the migration can
continue. To continue the repo (normally ~/.ipfs/
) needs to have
the following two directories:
blocks-v4
blocks-v5
If blocks-v4
does not exist but blocks-v5
does then it is likely
there was a problem towards the very end. To finish just create an
empty blocks-v4
directory. If neither directly exists but blocks
does then it could mean that there was a problem updating the repo
version file, but it could also mean that the migration failed to
start. If you are sure the migration ran than rename blocks
to
blocks-v5
and create an empty blocks-v4
directory and restart the
migration. If you are not sure about the state of the migration then
it is best to ask for help either on the #ipfs
irc channel or by
filing a bug report at https://github.com/ipfs/fs-repo-migrations.