On Sat, 28 Jan 2012 16:43:38 +0100 Christoph Anton Mitterer <[email protected]> wrote:
> Backuping the database on package upgrade fails with: > Preparing to replace ejabberd 2.1.9-1 (using .../ejabberd_2.1.9-1 > +b1_amd64.deb) ... Can't store backup in > "/var/backups/ejabberd-2012-01-28T16:36:54.mjvQ2n/ejabberd-database" > at node '[email protected]': I've looked at the code which does backup, and unfortunately this stack trace procuced by the Mnesia module > {'EXIT', {undef, [ > {mnesia_backup, open_write, > ["/var/backups/ejabberd-2012-01-28T16:36:54.mjvQ2n/ejabberd-database"]}, > {mnesia_log, safe_apply, 3}, > {mnesia_log, do_backup_master, 1}, > {mnesia_log, backup_master, 2}]}} is all what's reported back. >From this trace I can only tell that the (lightweight Erlang) process that hosted Mnesia died for *some* reason; there's no real display of a lower-level error, no POSIX errno number, no nothing. All I can tell that this problem has nothing to do with package upgrade per se. This most probably indicates a transient error (No space left on device? Permission problem? Something else? I dunno). Ideally, it would be cool to have more sensible error reporting, but this seems to require a change in the Erlang runtime. I won't close this bug as I'll try to communicate this with upstream or some Erlang expert. -- To UNSUBSCRIBE, email to [email protected] with a subject of "unsubscribe". Trouble? Contact [email protected]
