MME/HSS/eNB power-on race condition

[spencersevilla]

Complicated multi-part bug that we've seen in Bokondini with BaiCells Nova-233 eNB. Induced sometimes in the mornings when the entire system powers-on simultaneously at 0600.

1. HSS comes online, starts checking tables and keys. This takes a while if the table is large!!! (i.e. bug not seen for n=10; is seen for n=500).
2. MME comes online, starts the SCTP listener thread.
3. MME tries to connect to HSS using s6a; this fails. MME will attempt to reconnect in 30 more seconds.
4. HSS finishes table-checking and starts the s6a listener thread.
5. eNB initiates SCTP connection with MME; this fails since MME cannot yet speak to HSS.
6. MME connects to HSS using s6a. Once this connection succeeds, the MME is now in a stable state to receive eNB connections.

Observed reconnect behaviors:
MME: tries to connect to HSS every 30secs until successful connection occurs.
eNB: tries to connect to MME every 30secs. If unanswered, tries again. If answered/rejected, WILL NOT try again and must be rebooted.

Fix: Re-arrange the MME init code such that the SCTP listener thread (step 2) happens only after s6a and s1ap threads are successfully listening.