rocha
02/14/2025, 10:08 PMPraveen
02/15/2025, 9:34 PMrocha
02/16/2025, 5:31 PMrocha
02/16/2025, 5:31 PMrocha
02/16/2025, 5:32 PMrocha
02/16/2025, 5:32 PMrocha
02/16/2025, 5:32 PMPraveen
02/17/2025, 1:18 PMrocha
02/17/2025, 1:54 PMrocha
02/17/2025, 2:22 PMrocha
02/17/2025, 6:19 PMrocha
02/17/2025, 6:20 PMrocha
02/18/2025, 1:45 PMrocha
02/18/2025, 1:45 PMrocha
02/18/2025, 1:45 PMrocha
02/18/2025, 2:36 PMrocha
02/18/2025, 2:36 PMrocha
02/18/2025, 2:51 PMrocha
02/18/2025, 2:51 PMrocha
02/18/2025, 2:51 PMPraveen
02/19/2025, 9:46 AMPraveen
02/19/2025, 9:58 AM200 OK (INVITE)
confirms the SIP server accepted the call.
- Frame 8: ACK
sent in response to 200 OK
, confirming call establishment.
- No BYE
Message in the Capture
- A BYE
request is needed for a proper call termination, but it's missing from this trace which resulted into phone-call-provider-closed-webcoket-connection
- This suggests one of the following:
- The BYE
message wasn't captured.
- The call was terminated due to an issue outside of SIP signaling.
https://cdn.discordapp.com/attachments/1340082214146146435/1341710553416208475/rocha.pdf?ex=67b6fcd3&is=67b5ab53&hm=b3a159870624a2a67c0485689b05b755509b6b19ced9800fa3b1178ceae5de22&Praveen
02/19/2025, 10:03 AMrocha
02/19/2025, 1:37 PMrocha
02/19/2025, 1:38 PMPraveen
02/21/2025, 4:05 AMNir S (CEO/Founder @Cloudonix)
03/01/2025, 5:21 PMsip.vapi.ai
domain name. What I've noticed is that in some situations, the call is received at a specific gateway, then the SIP response comes from another server, which potentially just spun up and not yet has the active SIP dialogs populated to it. Again, this is only an assumption from my analysis.
At Cloudonix, we created our DNS resolver, so we maintain the signaling of sessions based on the first request, maintaining a statefull session. Technically speaking, this method is in direct contradiction to the SIP RFC, which is why Asterisk and RingCentral will have issues with that. Again, based on my analysis, this seems like a race condition occurring during servers spin up or scaling at Vapi, but this is just an assumption - I can't confirm that this is an issue.
BTW, if you wonder why I'm so familiar with Asterisk, look for my name in the source code 😀