#2802 closed defect (somebody else's problem)

startup failures non-obvious

Reported by: meejah Owned by: daira
Priority: normal Milestone: undecided
Component: code-frontend-magic-folder Version: 1.11.0
Keywords: Cc:
Launchpad Bug:


If you start up a node with a magic-folder and for some reason the shared-folder can't be accessed you'll get a long exception in logs/twistd.log ending with allmydata.mutable.common.UnrecoverableFileError: no recoverable versions.

However, the node keeps running but will not do any magic-folder "stuff" any more. No matter what, the error printed to the log should be something suitable for a user (e.g. "can't access magic-folder directory X" or so).

We should either:

  • kill the entire node if this happens
  • keep re-trying (and re-logging the error)

I have seen this occasionally when running check_magicfolder_smoke.py tests but am unsure how to reliably repeat it. The obvious thing of turning off all the storage servers and re-starting the Alice (or Bob) client doesn't appear to be sufficient.

Change History (2)

comment:1 Changed at 2016-08-08T22:30:35Z by meejah

Okay there's a way to repeat: run check_magicfolder_smoke.py, then run it with kill argv, then start just one of the storage nodes, then start Alice.

Alice will be when_connected_enough because there will be 2 storage servers connected (alice, plus the one storage node) so MagicFolder.ready is called, but this fails because the DMD can't be retrieved...

comment:2 Changed at 2020-06-30T13:45:05Z by exarkun

  • Resolution set to somebody else's problem
  • Status changed from new to closed

magic-folder has been split out into a separate project - https://github.com/leastauthority/magic-folder

Note: See TracTickets for help on using tickets.