[erlang-questions] unregistering a process during takeover

[Ulf Wiger]

Not sure based on what you’ve described. A possibility is that you’ve forgotten to start the application on b.

The next step, though, would be to also use a boot script. You could use rebar, relx, OTP systools or my personal pet project, setup [1].

Using a boot script, your applications are started automatically, or in the case of distributed apps, at least prepared to start, so that failover/takeover can work.

BR,
Ulf W

[1] http://github.com/uwiger/setup <http://github.com/uwiger/setup>
    https://github.com/uwiger/setup/wiki/Example%3A-generating-boot-scripts-from-a-running-system <https://github.com/uwiger/setup/wiki/Example:-generating-boot-scripts-from-a-running-system>


> On 23 Jan 2015, at 12:09, Nuku Ameyibor <nayibor@REDACTED> wrote:
> 
> hi ulf ,
> thanks.
> the takeover is working okay now  !!! .
> config a is this 
> 
> [{kernel,
>   [{distributed, [{m8ball,
>                    5000,
>                   [a@REDACTED, {b@REDACTED, c@REDACTED}]}]},
>    {sync_nodes_mandatory, [b@REDACTED, c@REDACTED]},
>    {sync_nodes_timeout, 30000}
>   ]
>  }
> ].
> 
> i start each node with  
> bin\werl.exe -sname c -config C:\wamp\www\erlang\proj\m8ball\apps\m8ball_1_0_0\config\c.config
> 
> there are 3 nodes .their conf files are similar with sync_nodes_mandatory being slightly different .
> 
> a small question .
> the first round of  failover and takeover seems to work fine .
> but after the first round (start node a,b,c  . kill a  b does failover .kill b c does failover . restart a,b  a does takeover from c killing c application on c .i restart shell on c  )
> 
>  after the first round when  i try to do a failover   onto b by killing a it doesnt work .
> was wondering how come since the first round of failover take over worked fine .
> 
> 
> 
> 
> On Fri, Jan 23, 2015 at 6:35 AM, Ulf Wiger <ulf@REDACTED <mailto:ulf@REDACTED>> wrote:
> 
>> On 23 Jan 2015, at 03:44, Nuku Ameyibor <nayibor@REDACTED <mailto:nayibor@REDACTED>> wrote:
>> 
>> =INFO REPORT==== 23-Jan-2015::02:20:57 ===
>> global: Name conflict terminating {m8ball_server,<2910.62.0>}
> 
> This suggests that you’re connecting the nodes *after* starting your applications, so that the application processes have time to globally register themselves first. Global detects the name conflict when trying to merge the registries of the two nodes.
> 
> You can for example use the ‘distributed’ option, described here:
> http://www.erlang.org/doc/apps/kernel/application.html#load-1 <http://www.erlang.org/doc/apps/kernel/application.html#load-1>
> 
> and here:
> 
> http://www.erlang.org/doc/man/kernel_app.html <http://www.erlang.org/doc/man/kernel_app.html>
> 
> BR,
> Ulf W
> 
> Ulf Wiger, Co-founder & Developer Advocate, Feuerlabs Inc.
> http://feuerlabs.com <http://feuerlabs.com/>
> 
> 
> 
> 

Ulf Wiger, Co-founder & Developer Advocate, Feuerlabs Inc.
http://feuerlabs.com



-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://erlang.org/pipermail/erlang-questions/attachments/20150123/fd43c610/attachment.htm>