NIP Publishing with two ADH servers using Elektron Java SDK

tbaker

I am writing a Java NI Provider to publish to our TREP infrastructure using the Elektron SDK. We have two ADH and two ADS servers. The ADH servers configured using hotStandby.

When testing I see that my subscribers don't see any data that is published to the standby ADH. When I failover the ADH servers I see that the previous standby server has stale data (the last data that was published directly to that server).

Should the NI Provider publish to just the active server? Or should it publish to the both at the same time? Multicast isn't currently supported by Elektron, but I could easily run the submit method on two publishers at once using RSSL for both.

I expected that the standby server has a mechanism to keep in sync with the active server. But that's not what I'm seeing. Perhaps I've misconfigured the data source?

This is similar to another question posted (https://community.developers.refinitiv.com/questions/10246/nip-publishing-on-2-adh-without-multicast.html). But it wasn't clear what is the reason to publish to both ADH servers? Should I just configure them to keep in sync?

Nipat Kunvutipongsak

Hello @tbaker,

From my understanding, the hot standby feature is used for making a fault tolerance on ADHs. So, if one of them goes down, the other once is still available and can continue providing services to downstream components.

If you expect for making the standby server performs data synchronization with the active server. Imagine the situation that there is something wrong with the active server, it will affect to the standby server as well, which means that the hot standby feature will be useless in this case.

Referring to your statement:

When testing I see that my subscribers don't see any data that is published to the standby ADH. When I failover the ADH servers I see that the previous standby server has stale data (the last data that was published directly to that server).

I'm not sure that you encountered this behavior from which type of topology, but in my test environment using topology 1, there were the scenarios as follows:

1. The active ADH went down.

2. The connecting ADS detected this problem and sent a DIRECTORY update message to the application with the status text: "A23: Source has gone down, requesting from another source."

3. The application will change the item status to 'Open/Suspect'.

4. Once the standby ADH became active, then ADS sent a DIRECTORY update to the application.

5. After that, ADS sent a new refresh message to the application again. The item status went back to Open/Ok.

That's why I'm curious about the different results. Please let me know if there is something that I'm overlooking.