Comment 1 for bug 1305758

Revision history for this message
Jeroen T. Vermeulen (jtv) wrote :

Got it!

The snapshot directory is created on the fly, as new files are downloaded from the repository. The import code also generates a JSON dump of the snapshot's metadata in memory. Then, it compares the JSON dump to what's in the current snapshot's maas.meta.

If the two are identical, there are no changes and further steps can be skipped. But if the two differ, the metadata has (probably) changed. In that case, the code writes the JSON dump to a new maas.meta file in the new snapshot, and promotes the snapshot to the current one.

The problem comes when the metadata has changed, but there are no new downloads. The code attempts to write maas.meta, but the snapshot directory does not exist!