I made a backup with 'ejabberdctl node@host backup '.
Now i want to restore this backup with 'ejabberdctl node@host restore ', but all i get is:
RPC failed on the node ejabberd@hhserver: {'EXIT',
{{case_clause,
{aborted,
{no_exists,presence}}},
[{ejabberd_ctl,process,1},
{rpc,
'-handle_call/3-fun-0-',
5}]}}
Help!!!
It seems that ejabberd (or
It seems that ejabberd (or Erlang/Mnesia) habe Problems if you try to restore a backup wich was made if some of your loaded modules don't exist anymore or have changed (by upgrades to later versions). The backups (made by 1.1.1), by the way, where incomplete (contained only half of all the users). Luckily I had some older backups (from Version 1.0.0), which restored successfully (but I had to downgrade to 1.0.0 to make them work).
So, is seems you cannot upgrade versions if you want to make functioning backups.
Well, that seems to be a very SERIOUS bug?
Thank you for letting us
Thank you for letting us know.
We are investigating this problem.
--
Mickaël Rémond
Restore to 1.1.1
Hello,
I have found where the problem comes from. The restore should now be possible even if modules are no more used or if unknown tables exists in it.
The modified ejabberd_ctl.erl is available from:
Please, could you let me know if this solves your problem ?
--
Mickaël Rémond
Works!
Yes, it works as it schould! Many thanks!
Was that also the reason why the backup itselb was incomplete? Test-backupped the restored database and this time the backup seems to be complete.
Backup
Hello,
To be frank, I did not manage yet to reproduce a partial backup. There is a single Mnesia command that take care of the whole backup. If there is a problem, an error is return. So for now, I cannot imagine cases where it could only be partially done (apart from disk space limit).
--
Mickaël Rémond
Very weird... I made 4
Very weird... I made 4 backups (all in a row) and they were all incomplete. And there was plenty of disk space. But as for now. everything seems to be complete...
How did you check your
How did you check your backup ?
Did you restore it on a completly new install, or an install with already data in it ?
--
Mickaël Rémond
I checked it by restoring it
I checked it by restoring to a new test-install (deleted the spool-directory before restoring) and by browsing it with a text/hexeditor. I found only a part of all users. I didn't got into it further by analysing if there's something special about the included or missing users, by I think the missing ones where the "youngest" (by registration date), but I'm really not sure.
I also tried to "replay" the upgradepath (restored the db into a 1.0.0 and upgraded step by step to 1.1.1), but no trace of the problem. So I guess it was due to weird circumstances or Murphy had something to do with it ;-)
If you simply checked with a
If you simply checked with a text/hexeditor you might have missed some users that was actually in the database. For example, there is a log file where new inserts are temporary stored. You might have missed them.
I think the database might actually been complete.
--
Mickaël Rémond
Nono, it's realyy incomplete
Nono, its really incomplete. I forgot to mention that it also has only about half the size of the backup I made some days before. Also I searched the backup for some of the missing users after I restored it. But however, the problem seems to have disappeared.