OpenSRF Service Listeners should be more forgiving of missing routers
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
OpenSRF |
New
|
Undecided
|
Unassigned |
Bug Description
Similar to bug 1954519 individual service listeners should be more robust in the face of network unpleasantness between them and their routers, ideally automatically re-registering with their routers once reconnected.
An important point of differentiation is that C services currently just die if the routers aren't already running at service startup; perl services start and wait patiently to be told to re-register.
Testing shows that both C and perl running services survive having their routers killed and restarted* and the overall system seems to operate normally once services have been re-registered. This is similar to what would happen in a multi-machine brick if the machine hosting the routers had connectivity issues.
* Looks like we have nothing like --start-routers and --stop-routers; they're only started / stopped with *_all.