Looking for ejabberd docs?

To access the most up-to-date ejabberd documentation, please visit docs.ejabberd.im »

Failover HA behaviour

Submitted by S.T.E.N.C.H. on Mon, 2007-07-02 12:10

Hi,

I would like to know where I can read up about the behaviour and expected HA on failover, e.g. will sessions be maintained if a cluster node goes down? What is the expected behaviour of messages in flight? etc.

Thanks in advance

ejabberd nodes in a cluster

Submitted by mfoss on Fri, 2007-07-06 11:06.

ejabberd nodes in a cluster can share some tables of the database: accounts (username and password), roster, offline messages, vcard...

Each node has his own client and server connections. When a node crashes, all his connections are lost.

I don't know what will happen with messages in the flight.

Thanks - seems like I will

Submitted by S.T.E.N.C.H. on Tue, 2007-07-17 13:48.

Thanks - seems like I will have to make my own arrangements!

What do you mean ? The

Submitted by mremond on Tue, 2007-07-17 15:38.

What do you mean ?

The cluster in ejabberd is top notch and I doubt you can do better with some quick hack ...

--
Mickaël Rémond
Process-one

Clarification Question on ejabberd Clustering

Submitted by TFerguson on Fri, 2008-05-16 17:20.

I have two machines set up in an ejabberd cluster, and everything works as advertised when they are both online. However, if the first (primary) node is offline, I cannot log in to my ejabberd domain via the second (secondary) node. Is this the intended behavior, or should secondary nodes still be able to operate (as an ejabberd server) in the event the primary node is down? When I look at the databases for each node in the Web Admin, they are identical for each node, which I assumed meant they were operating independently, but with replication of the tables. It's like it is looking for something for authentication that only exists on the primary server??

If this is not the intended behavior, are there any clues for where I can begin to look for misconfiguration? I don't see anything that looks out of order in the logs, etc. I have no hacks to my setup, just used the standard config with the clustering setup instructions in the doc.

TIA -TF

*******************************
ejabberd 2.0.0 on OpenSuSE 10.2

The table 'passwd' must be replicated in each node.

Submitted by mfoss on Sun, 2008-05-18 17:49.

TFerguson wrote:

if the first (primary) node is offline, I cannot log in to my ejabberd domain via the second (secondary) node. Is this the intended behavior, or should secondary nodes still be able to operate (as an ejabberd server) in the event the primary node is down?

The ejabberd Guide says:

Now you can add replicas of various tables to this node with `mnesia:add_table_copy' or `mnesia:change_table_copy_type' as above (just replace `schema' with another table name and `disc_copies' can be replaced with `ram_copies' or `disc_only_copies').

Which tables to replicate is very dependant on your needs, you can get some hints from the command `mnesia:info().', by looking at the size of tables and the default storage type for each table on 'first'.

Replicating a table makes lookups in this table faster on this node. Writing, on the other hand, will be slower. And of course if machine with one of the replicas is down, other replicas will be used.

If you didn't instruct Mnesia @ node2 to replicate the table 'passwd' from node1, then you can only login in node2 when node1 is available.

If you already configured Mnesia @ node2 to replicate the table 'passwd' from node1, then I don't have clues of where is the problem.