https://retellai.com/?utm_source=discord&utm_medium=seo logo
Powered by
Call session not ending
# ❓・bug-report
h
We have an urgent situation where we can’t see the call transcript as the call is showing as ongoing and we need that transcript. The client CEO called my cell and confirmed call is complete, but API still shows it as ongoing and has exceeded the 2 hour cutoff time Call ID: call_7de6763b5fbe06092744f3a7b98
s
This is a very important call, we need help! The call is still in progress and no APIs are working
> 1. Does your system have the raw audio for this call, even if unprocessed? > 2. Does the Twilio SID have a recording on your Twilio account? > 3. Can you manually transition this call to "ended" so the transcript and recording are generated?
s
Hi @harii
What is your agent id
h
agent_d3ce7ca9bcc70f803dbba12820
s
Ok I report this to team to look
h
@Shehroze -Support @Retell any update? We need that data today
s
@harii Team is looking into it
s
We had the SID from twillio too, would that be needed?
h
@Shehroze -Support @Retell can you get us any of the data?
s
Hi @harii this call session is terminated due to a instance draining issue, we are not able to retrieve data.
h
@Shehroze -Support @Retell Wow, this is a major problem. What is to prevent these issues from happening?? What can we do to prevent this
By “instance draining” you mean the server was being taken offline during the call and so it lost everything?? What is the frequency
This was a 1 hour call with the CEO of a 400 person company
Is this on Retell side or on your infra provider? I need to manage fallout
What can we do to ensure there’s always a backup of a call? Something we insert programmatically via API? What is your recommendation
s
:/ why is the data not retrievable? It was a 60 min call
@Shehroze -Support @Retell we need an RCA, what happened here, this was a very very important call.
s
Hi @Shweta_halo It is a constraint on the Retell infrastructure side. In rare cases, an instance may enter a draining state while processing calls, with a maximum draining duration of 1 hour currently. We plan to release a patch to ensure data is uploaded before instance shutdown. However, please note that in such rare cases, calls may be terminated before reaching the maximum call duration.
s
When you say call is terminated, does that mean ended or deleted? because if the call was ended, which it should have been, why did we lose the conversation??
Doesnt Twilio track it? with the SID on their end?
s
No, Twilio does not store call data; it only maintains logs related to call connectivity. When call ended, typically we will save the session data into our database and cloud storage, since this call session cut off in the middle, our instance did not have then chance to save them.
h
What's the expected frequency of this happening? It would concretely help us if we can tell the customer, "This happens to 0.02% of calls" or something. This was in the middle of the afternoon PT
s
This is a huge risk for us and our customers (CEOs of mid market companies whose time is super valuable), you are saying this can happen anytime and we might have to just let go of that data?!
s
Hi @harii To be honest, it's really rare, about 0.0003%.
h
This instance draining happened to 2 of our calls - both the one that was not saved and one that finished 5 min before it, where it appeared the agent ghosted the caller. Both started within a minute of each other. Do you enable something to ensure the same connection isn’t used twice, so multiple calls aren’t affected by the server going down? Or when will this patch be released?
s
It's very rare, happened with you and Teams plan is to release a patch to ensure data is uploaded before instance shutdown.
h
Yes but when will this patch be released?
s
We’ve identified the issue and the team will work on a fix. A patch should be released soon.
h
@Shehroze -Support @Retell has the patch been released?
PreviousNext
Powered by