WARNING - still work in progress!!!

Debian 6.0 Squeeze, however I think this doesn't depend on your GNU/Linux distribution.

• Migrating from: ejabberd-2.0.5-x86_84
• Migrating to: ejabberd-2.1.5-3+squeeze1

I recommend to backup /var/lib/ejabberd prior to any changes. If anything goes wrong or doesn't work, you can start very quickly from scratch.

root@foo:~# # backup
root@foo:~# cp -apr /var/lib/ejabberd /tmp

root@foo:~# #restore, which is just way around
root@foo:~# cp -apr /tmp/ejabberd /var/lib/

• dump database at old ejabberd

root@foo:~# ejabberdctl backup /tmp/ejabberd.backup

• move backup, configs etc. to new machine, if needed, using eg. SCP
• change hostname, if needed or got changed, as described in Migrate FAQ
• restore database as shown bellow. Skipping tables can be iterative process:

root@foo:~# sh /etc/init.d/ejabberd stop
root@foo:~# ejabberd debug
[...]
(ejabberd@foo)1> mnesia:restore("/tmp/ejabberd2.backup",[{skip_tables,[disco_publish]}]).
{aborted,{no_exists,user_caps}}
(ejabberd@foo)2> mnesia:restore("/tmp/ejabberd2.backup",[{skip_tables,[disco_publish,user_caps]}]). 
{aborted,{no_exists,user_caps_resources}}
(ejabberd@foo)3> mnesia:restore("/tmp/ejabberd2.backup",[{skip_tables,[disco_publish,user_caps,user_caps_resources]}]).
{atomic,[config,privacy,local_config,passwd,irc_custom,
         roster,last_activity,sr_user,offline_msg,route,motd,acl,s2s,
         vcard,caps_features,sr_group,mod_register_ip,vcard_search,
         motd_users,session,private_storage,pubsub_item,muc_room,
         pubsub_state,iq_response,muc_registered,muc_online_room|...]}
(ejabberd@foo)4>q().

• start ejabberd

root@foo:~# sh /etc/init.d/ejabberd start

• now, I would like to say “enjoy”, but there was no joy. I couldn't get authenticated no matter what. Perhaps password encryption got changed, I don't know.

• dump database at old server

root@foo:~# ejabberctl dump /tmp/ejabberd.dump

• move backup, configs etc. to new machine, if needed, using eg. SCP
• start ejabberd at new server
• restore database
  • you may try to restore database right away by following command. However, it didn't work for me, not even after removing disco stuff.
  • OR write a parser to get rid of problematic things in .dump
    • problematic stuff is:
      • everything related to disco
      • user_caps - since it has a lot of entries, I didn't want to remove it by hand. Thus I've removed just 'mod_caps'
      • pubsub_node - I had to remove data related to this table, because something got changed and no, I'm not going to study Erlang right now

root@foo:~# ejabberdctl load /tmp/ejabberd.dump

• hopefully enjoy!