TA900 with voice users not generating outbound REGISTERs - some kind of retry timeout bug?
We have a couple hundred TA900’s (we mostly see this issue with 904s and 908s) deployed with varying numbers of FXS/Emulated POTs/Voice Users (whatever you want to call them) configured to register back to individual subscribers provisioned on our softswitch.
As we have scaled up the number of TA900s we have deployed, we have noticed an issue where periodically some of the Voice Users on the devices are in an unregistered state, and the TA900 is making no attempt to send outbound registers (observed via '#debug sip stack messages').
We have also observed that when a TA900 is in this state, running a '#sip trunk-registration force-register' does not seem prompt the TA900 to generate an outbound REGISTER for the voice users in question - and generally you have to remove those voice users and then re-add them to get the TA900 to send the REGISTER.
Adtran support keeps asking us for a live debug of this, but obviously we don’t generally know when this is going to happen and we haven’t cooked up a way to reproduce, so being able to provide that seems like a bit of a longshot.
We thought maybe the sip stack was being swamped by SIPvicious type internet SIP bots, so we rolled out SIP ACLs, but that theory appears to be busted as we have observed voice users on devices with SIP ACLs fall into this state.