Failover possible?

[madkote]

trafficstars

hi,

is there failover functionality?

E.g.: https://github.com/jasonrbriggs/stomp.py/blob/010b245bc869381e25895b26d987ca19df1d2a2c/stomp/connect.py#L123

[pedrokiefer]

It supports automatic reconnecting, but there is no support for a list of servers. I think it's simple to extend the reconnect code add this functionality though.

[madkote]

thx for reply.

list of servers would be very useful.

[madkote]

it is a loop over tuples of hosts and ports, instead connecting to one host+port.

[pedrokiefer]

Maybe extracting the reconnect logic to another class, and make it composable, so one can have different heuristics for reconnection, like:

• retry forever on one server (default)
• try first server, failover to second server (failover strategy)
• try first server, retry once, try second server, retry once (failover with retry strategy)
• ...

Sounds good to you ?

[madkote]

Hi @pedrokiefer , thank for your reply.

• retry forever on one server -> should be customizable -1, .., n == infinite to N
• failover -> yes, it sounds good. There should be then two retry values - for retry on server and retry failover. Whereby, I personally, would do only one attempt to one of the servers and go to the next in case of failure -> interest here is to find a server asap.

Are you going to implement this soon? otherwise I can also leave with my customization.

[madkote]

when implementing failover be aware of following behaviour with ActiveMQ in cluster mode:

• connections are only acceptable to the master
• when connection to a slave, the connection is accepted, but slave disconnects it after some short period of time.