[erlang-questions] How to dynamically add one node to erlang's distrubted application?
Wed Sep 10 15:03:29 CEST 2008
2008/9/10 devdoer bird <>
> Thanks. If I have 100 machines ,I have to edt the config files on all 100
> machines, that's unacceptable. Are there any solutions?
I don't have an answer to your specific question, but...
I wouldn't recommend using this feature across 100 machines. Do you really
need an application to failover across 100 machines? I had trouble with this
feature even across four erlang nodes. The nodes used to hang arbitrarily at
various times. It was a long time ago and lots of fixes have been made for
this but I'm still quite wary of it. I've found that designing my
applications so that failover is handled at the application level gives me
If you are using this feature to ensure that only one instance of your
application is running at any one time, then you are getting a false sense
of security. One node can never definitely know if it's peer is running or
not. Network outages between the two nodes can make it look like the other
node is down, when in reality they are both running.
> 2008/9/10 Nicolas Charpentier <>
>> devdoer bird wrote:
>>> Now I want to add one more node to the system .How can I do this using
>>> erlang's distributed applicaiton mechanism? Do I have to edit the config
>>> file on all 3 nodes to add the 4th node and restart all of them?
>> if you can't stop your nodes because others mandatory applications are
>> running you can try to use application:load/2 on all your nodes. The second
>> param is the distribution parameter which overload kernel value.
>> To be able to load an application, it have to be stopped.
>> Don't forget to update the config file for the next node restart.
>> I never tried to do that but the documentation seems to say that it's
> erlang-questions mailing list
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the erlang-questions