HA regression with active sessions on 3.12.2.2

[dudley.aoi]

1. Brekeke Product Name and Version:
Brekeke SIP Proxy Server 3.12.2.2

2. Java version:
11 (required for 3.12.2.2, whereas Java 17 worked perfectly with 3.12.1.5)

3. OS type and the version:
RHEL 7.9

4. UA (phone), gateway or other hardware/software involved:
N/A (tested with Avaya and various other SIP implementations)

5. Your problem:
Setup: Two Brekeke SIP Proxy Servers running with mirroring/redundancy for HA.

Event: One of the proxy servers is rebooted (reboot or power failure, etc).

Observed outcome:
Ongoing/active calls remain up, however no longer appear under Active Sessions on the rebooted/restarted peer. Active calls do still show normally on the peer that was not rebooted/power cycled.

Expected outcome:
Ongoing/active calls remain up and are listed in Active Sessions when the peer rejoins the mirror. Ongoing calls show under Active Sessions on both peers.

Additional details:
This does not appear to impact switching roles manually using "Switch the role" button on either the primary or secondary in our testing. Calls no longer showing under Active Sessions only occurs after reboot of either member while there are ongoing sessions. If the designated secondary is then rebooted, the call will still remain active, but then will not show under Active Sessions on either mirror member. So long as the designated secondary stays up, the ongoing calls will still show as Active Sessions on the designated secondary (which is then acting as either standalone or primary). New calls appear normally; this only impacts ongoing sessions during failover resulting from reboot/power cycling.

This does not occur under 3.12.1.5 with identical setup (pre-upgrade to 3.12.2.2). Note that 3.12.2.2 will not allow login without backversioning Java to Java 11; 3.12.2.2 does appear to start, but the web console will not work under Java 17 as it did with 3.12.1.5.

[dudley.aoi]

Clarification:
Switching the roles manually after the event has no effect on the situation whatsoever: if the call originated with both primary and secondary up, and the designated primary is rebooted, then when the designated primary comes back up, there is a switch role performed (on either the secondary or primary); the ongoing call will still NOT show up as an active session on the designated primary. It will still show as an active session on the designated secondary, until that is rebooted as well. Then the call will still remain active, but won’t show in active sessions anywhere.

The roles can be switched back and forth, the calls will continue, but will no longer show as active sessions on whichever peer was rebooted/restarted. As long as one peer remains, sessions for the cluster can still be found on that particular peer.

[dudley.aoi]

It appears there might be a log buffering regression there too, as even with a graceful reboot sometimes the call might not appear in the logs of the designated primary. That's really odd.

We'll be staying on 3.12.1.5 until this bug gets fixed.

[Mike]

Ongoing/active calls remain up, however no longer appear under Active Sessions on the rebooted/restarted peer. Active calls do still show normally on the peer that was not rebooted/power cycled.

It is what designed. I added your request in the TODO.

[dudley.aoi]

This appears to be fixed in 3.13.0.0 (however it still requires going back to Java 11).