Manual Failover

Directed here by a Senior Technical Specialist from Refinitiv ticket: 7790900

Having discussed this with our Mutual Client we wanted to get some more insight into the issue reported in 7790900.

Because the connection to the EED didn't appear to have failed we didn’t failover which meant that we didn't get any Market Data due to an issue with the EED. We had to manually configure this to point to another EED in order to get that application to work as expected with an overnight change.

From what was stated by the Technical Specialist, it is our impression that there is no way to manually failover to another EED in a serverList without having the restart our application with a edited configuration as this would “shut down connectivity to the affected EDGE”, is there no other functionality via an API to do this failover Manually without having to stop our application?

Best Answer

  • nick.zincone
    Answer ✓

    Hi @akyle,

    In order for the library you are using to automatically failover, a specific event must occur that it can detect. In this case, the event is a disconnect or the library believes it has lost it connection due to no heartbeats - see this thread for further details. If neither one of these events occur, the library considers the connection ok - as you indicated.

    Consumers are always connected to an infrastructure that contains a backend that manages issues automatically - unlike the situation you are describing where you are connected directly to an EED. Because of this, you are losing that added benefit the backend brings by having multiple, active connections and being able to rollover if there are connection issues or service going down issues etc. That being said, your software will have to programmatically decide when it thinks there are issues if the connection is active and alive. Both service status and market data item status messages will be sent that you can potentially use to decide to forcibly break the connection.

Answers

  • Hi @akyle

    I had a quick read of ticket 7790900 and this relates to a hardware failure / replacement scenario. Also, it appears that there is some middleware involved in the mix - which may be connecting directly to the EED. Are you the developer of this middleware?

    Can you confirm a few details:

    • Is your application connecting directly to an EED? It is very rare for an application to connect directly to an EED.
    • Which API (and programming language) are you using?
    • Did you have a serverList (RFA) or ChannelSet (EMA) in your config with multiple servers (EED or ADS) specified - in order for the API to try an alternative server?
    • Do you have Warm Standby configured at your application level?
    • Whereabouts in your system were you expecting the failover to take place - which component?