Calls dropped by RTP inactivity and media timeouts

See also

Sources

1. 1
   The Asterisk res_pjsip endpoint option rtp_keepalive makes Asterisk send an RTP comfort noise frame at the configured interval, default 0, which the docs describe as useful behind NAT or a firewall.

   res_pjsip - Asterisk Documentation · endpoint object > rtp_keepalive · Checked 2026-10-14
2. 2
   Asterisk res_pjsip has a separate endpoint option rtp_timeout_hold that sets the seconds without RTP while the call is on hold before the channel is considered dead, default 0.

   res_pjsip - Asterisk Documentation · endpoint object > rtp_timeout_hold · Checked 2026-10-14
3. 3
   In Asterisk res_pjsip the endpoint option rtp_timeout sets the seconds without RTP (while off hold) before a channel is considered dead, and its default 0 means no check.

   res_pjsip - Asterisk Documentation · endpoint object > rtp_timeout · Checked 2026-10-14
4. 4
   The Asterisk res_pjsip endpoint options rtp_timeout, rtp_timeout_hold and rtp_keepalive are each documented as available since Asterisk 13.5.0.

   res_pjsip - Asterisk Documentation · endpoint object > rtp_timeout / rtp_timeout_hold / rtp_keepalive, Since field · Checked 2026-10-14
5. 5
   On AudioCodes Mediant SBCs (User Manual 7.60) Broken Connection Timeout [BrokenConnectionEventTimeout] is the interval, in 100-msec units, after which an established call is disconnected if RTP packets are not received, valid range 3 to about 2684354.

   Answer and Disconnect Supervision Parameters · Answer and Disconnect Supervision Parameters > Broken Connection Timeout · Checked 2026-10-14
6. 6
   The AudioCodes 7.60 manual prints the Broken Connection Timeout default as 100 msec while defining the parameter in 100-msec units, so the effective default (0.1 s or 10 s) cannot be settled from the page alone; 10 s is the more plausible reading.inferred

   Answer and Disconnect Supervision Parameters · Answer and Disconnect Supervision Parameters > Broken Connection Timeout, default sentence · Checked 2026-10-14
7. 7
   AudioCodes broken-connection handling can be configured per IP Profile, and when configured on an IP Profile the device ignores the global parameter for calls associated with that profile.

   Answer and Disconnect Supervision Parameters · Answer and Disconnect Supervision Parameters > Broken Connection Timeout, IP Profile note · Checked 2026-10-14
8. 8
   AudioCodes No RTP Detection Timeout [NoRTPDetectionTimeout] is set in msec, range 0 to 50000, and its default 0 disables disconnecting or re-routing calls for which RTP is never received.

   Answer and Disconnect Supervision Parameters · Answer and Disconnect Supervision Parameters > No RTP Detection Timeout · Checked 2026-10-14
9. 9
   AudioCodes states that for the broken-connection disconnect feature to function, silence suppression must also be disabled.

   Answer and Disconnect Supervision Parameters · Answer and Disconnect Supervision Parameters > Broken Connection Timeout, note · Checked 2026-10-14
10. 10
    On Cisco IOS XE media-inactivity detection is performed using RTP by default, and media-inactivity-criteria selects all, rtcp, rtp or rtplib detection.

    Gateway Configurations - IOS-XE VoIP · Gateway Configurations > media-inactivity-criteria · Checked 2026-10-14
11. 11
    With both documented defaults (timer receive-rtcp 5 and ip rtcp report interval 5000 ms) the Cisco IOS RTCP-based media inactivity period works out to about 25 seconds.inferred

    SIP Configuration Guide, Cisco IOS Release 15M&T - Configuring SIP QoS Features · Configuring the SIP Media Inactivity Timer, both command steps · Checked 2026-10-14
12. 12
    Cisco IOS recognises hold for the SIP hold timer by c=0.0.0.0 or by a=sendonly or a=inactive in the SDP.

    SIP Configuration Guide, Cisco IOS Release 15M&T - Configuring SIP QoS Features · Information About SIP QoS > SIP Hold Timer Support · Checked 2026-10-14
13. 13
    The Cisco IOS SIP hold timer (timers hold) terminates calls held longer than a configurable 15 to 2880 minutes, default 2880.

    SIP Configuration Guide, Cisco IOS Release 15M&T - Configuring SIP QoS Features · SIP Hold Timer Support; Configuring the SIP Hold Timer · Checked 2026-10-14
14. 14
    When the Cisco IOS SIP Media Inactivity Timer expires the gateway sends a SIP BYE and sends a Q.931 DISCONNECT toward TDM with cause code value 3 (no route).

    SIP Configuration Guide, Cisco IOS Release 15M&T - Configuring SIP QoS Features · Information About SIP QoS > SIP Media Inactivity Timer · Checked 2026-10-14
15. 15
    On Cisco IOS the media inactivity period is the timer receive-rtcp multiplier times the ip rtcp report interval, and the call is disconnected if no RTCP arrives in that period.

    SIP Configuration Guide, Cisco IOS Release 15M&T - Configuring SIP QoS Features · Configuring the SIP Media Inactivity Timer, timer receive-rtcp step · Checked 2026-10-14
16. 16
    Cisco IOS documents that the media inactivity timer is deactivated when a call is placed on hold and no media packets are sent, which is why a separate SIP hold timer exists.

    SIP Configuration Guide, Cisco IOS Release 15M&T - Configuring SIP QoS Features · Information About SIP QoS > SIP Hold Timer Support · Checked 2026-10-14
17. 17
    The Cisco IOS SIP Media Inactivity Timer lets a gateway monitor and disconnect VoIP calls if no RTCP packets are received within a configurable period, to release hung sessions and their DSP and TDM resources.

    SIP Configuration Guide, Cisco IOS Release 15M&T - Configuring SIP QoS Features · Information About SIP QoS > SIP Media Inactivity Timer · Checked 2026-10-14
18. 18
    The Cisco IOS SIP Media Inactivity Timer requires both ip rtcp report interval and timer receive-rtcp so that the gateway uses RTCP detection rather than RTP detection.

    SIP Configuration Guide, Cisco IOS Release 15M&T - Configuring SIP QoS Features · How to Configure SIP QoS Features > Configuring the SIP Media Inactivity Timer · Checked 2026-10-14
19. 19
    The Cisco IOS 15M&T SIP configuration guide gives the timer receive-rtcp multiplier range as 2 to 1000 with a default of 5.disputed

    SIP Configuration Guide, Cisco IOS Release 15M&T - Configuring SIP QoS Features · Configuring the SIP Media Inactivity Timer, timer receive-rtcp step · Checked 2026-10-14
20. 20
    The Cisco DevNet IOS XE VoIP gateway configuration page gives the timer receive-rtcp multiplier range as 4 to 1000 with a recommended value of 5.disputed

    Gateway Configurations - IOS-XE VoIP · Gateway Configurations > timer receive-rtcp · Checked 2026-10-14
21. 21
    On Cisco IOS ip rtcp report interval accepts 1 to 65535 milliseconds with a default of 5000.

    SIP Configuration Guide, Cisco IOS Release 15M&T - Configuring SIP QoS Features · Configuring the SIP Media Inactivity Timer, ip rtcp report interval step · Checked 2026-10-14
22. 22
    On Cisco IOS XE timer media-inactive enables media inactivity detection using the DSP, with a range of 4 to 1000 and a recommended media-trans-interval of 5.

    Gateway Configurations - IOS-XE VoIP · Gateway Configurations > timer media-inactive · Checked 2026-10-14
23. 23
    Cisco DevNet lists the media-inactivity-criteria and timer YANG models as supported from Cisco IOS XE 17.7.1.

    Gateway Configurations - IOS-XE VoIP · Gateway Configurations, 17.7.1 YANG list · Checked 2026-10-14
24. 24
    Drop timing separates causes: drops at a fixed short interval after hold or mute point to an RTP-only inactivity timer, drops near half or all of a Session-Expires value point to session refresh, and long fixed drops on hold point to hold timers such as Cisco timers hold or a silent-timeout.inferred

    Issues that affect inbound Direct Routing calls · Calls drop after a specific duration (read with Cisco SIP Hold Timer and rtpengine --silent-timeout claims) · Checked 2026-10-14
25. 25
    A FreeSWITCH media timeout hangs up the call with hangup cause SWITCH_CAUSE_MEDIA_TIMEOUT.

    Chapter 17: Media Handling (FreeSWITCH documentation) · Media timeout hangup cause paragraph · Checked 2026-10-14
26. 26
    In FreeSWITCH media_hold_timeout sets the milliseconds of RTP inactivity allowed while on hold, and if it is unset it defaults to ten times media_timeout.

    Chapter 17: Media Handling (FreeSWITCH documentation) · Media timeouts: channel variables (media_hold_timeout) · Checked 2026-10-14
27. 27
    In FreeSWITCH the preferred RTP inactivity control is the per-channel variable media_timeout, expressed in milliseconds, where 0 disables the timeout.

    Chapter 17: Media Handling (FreeSWITCH documentation) · Media timeouts: channel variables (media_timeout) · Checked 2026-10-14
28. 28
    The FreeSWITCH Sofia profile parameters rtp-timeout-sec and rtp-hold-timeout-sec (in seconds) are deprecated and FreeSWITCH logs a warning when they are parsed at profile load.

    Chapter 17: Media Handling (FreeSWITCH documentation) · Deprecated profile parameters · Checked 2026-10-14
29. 29
    The FreeSWITCH media handling chapter gives vanilla configuration values of 300 seconds for rtp-timeout-sec and 1800 seconds for rtp-hold-timeout-sec.

    Chapter 17: Media Handling (FreeSWITCH documentation) · Deprecated profile parameters · Checked 2026-10-14
30. 30
    When the media flow timer sequence expires, Oracle Communications SBC sends a BYE carrying a Reason header indicating the session was cleared due to media flow guard timer expiry.

    Per-Realm Media Guard Timers · Per-Realm Media Guard Timers, expiry paragraph · Checked 2026-10-14
31. 31
    Oracle Communications SBC option disable-guard-timer-sendonly, set in media-manager-config, disables the initial and subsequent media guard timers when a SIP call is put on hold via c=0.0.0.0, a=inactive or a=sendonly.

    SIP Disable Media Inactivity Timer for Calls Placed on Hold · SIP Disable Media Inactivity Timer for Calls Placed on Hold, body and CLI example · Checked 2026-10-14
32. 32
    Oracle documents that enabling disable-guard-timer-sendonly also disables the guard timers for calls that are not on hold but are one-way audio applications.

    SIP Disable Media Inactivity Timer for Calls Placed on Hold · SIP Disable Media Inactivity Timer for Calls Placed on Hold, note paragraph · Checked 2026-10-14
33. 33
    On Oracle Communications SBC the media-manager-config initial-guard-timer is the seconds allowed before the first packet of a dynamic flow arrives, default 300, and 0 disables it.

    media manager config · media-manager-config > initial-guard-timer · Checked 2026-10-14
34. 34
    On Oracle Communications SBC the hold option must be added as options +disable-guard-timer-sendonly, because typing the option without the plus sign overwrites previously configured options.

    SIP Disable Media Inactivity Timer for Calls Placed on Hold · SIP Disable Media Inactivity Timer for Calls Placed on Hold, configuration steps · Checked 2026-10-14
35. 35
    When a flow crosses two realms with per-realm guard timers, Oracle Communications SBC uses the lower of the ingress and egress settings.

    Per-Realm Media Guard Timers · Per-Realm Media Guard Timers, ingress/egress paragraph · Checked 2026-10-14
36. 36
    On Oracle Communications SBC a realm-config guard timer value of -1 (the default) falls back to the global media-manager timers, while 0 disables that timer for all flows traversing the realm.

    Per-Realm Media Guard Timers · Per-Realm Media Guard Timers, value semantics paragraph · Checked 2026-10-14
37. 37
    On Oracle Communications SBC the media-manager-config subsq-guard-timer is the maximum seconds allowed between packets in a dynamic flow, default 300, and the SBC notifies the signaling application when it is exceeded.

    media manager config · media-manager-config > subsq-guard-timer · Checked 2026-10-14
38. 38
    Oracle Communications SBC has separate tcp-initial-guard-timer and tcp-subsq-guard-timer parameters for media-over-TCP flows, each defaulting to 300 seconds.

    media manager config · media-manager-config > tcp-initial-guard-timer and tcp-subsq-guard-timer · Checked 2026-10-14
39. 39
    Under RFC 3264 a sendrecv media stream is placed on hold by marking it sendonly, and a recvonly stream is placed on hold by marking it inactive.

    RFC 3264 — An Offer/Answer Model with the Session Description Protocol (SDP) · Section 8.4 Putting a Unicast Media Stream on Hold · Checked 2026-10-14
40. 40
    RFC 3264 states that RTCP is still sent and received for sendonly, recvonly and inactive streams, so a held call can still carry RTCP even when RTP stops in one or both directions.

    RFC 3264 — An Offer/Answer Model with the Session Description Protocol (SDP) · Section 5.1 Unicast Streams · Checked 2026-10-14
41. 41
    RFC 3550 recommends a fixed minimum interval of 5 seconds between compound RTCP packets, and the fixed minimum should still be used for timeout calculations even where a reduced minimum is allowed.

    RTP: A Transport Protocol for Real-Time Applications · Section 6.2 RTCP Transmission Interval · Checked 2026-10-14
42. 42
    With RFC 3550 defaults (5 second minimum interval and multiplier 5) an RTCP-based liveness check that follows the RFC would not time out a silent-but-RTCP-sending peer and would need roughly 25 seconds of total RTP and RTCP silence before declaring a peer gone.inferred

    RTP: A Transport Protocol for Real-Time Applications · Sections 6.2 and 6.3.5 combined · Checked 2026-10-14
43. 43
    RFC 3550 times out a session member that has sent neither RTP nor RTCP for M times the deterministic RTCP interval Td, with the multiplier M defaulting to 5.

    RTP: A Transport Protocol for Real-Time Applications · Section 6.3.5 Timing Out an SSRC; Section 6.2.1 · Checked 2026-10-14
44. 44
    RFC 6263 says an RTP/RTCP agent behind NAT should send keepalive traffic regardless of whether the media stream is inactive, sendonly, recvonly or sendrecv, because hold can let NAT bindings expire.

    Application Mechanism for Keeping Alive the NAT Mappings Associated with RTP / RTP Control Protocol (RTCP) Flows · Section 1 Introduction (as reported by fetch; recheck section number) · Checked 2026-10-14
45. 45
    RFC 6263 recommends RTCP multiplexed with RTP as the keepalive mechanism and does not recommend the other listed mechanisms such as empty packets or comfort-noise RTP.

    Application Mechanism for Keeping Alive the NAT Mappings Associated with RTP / RTP Control Protocol (RTCP) Flows · Section 5 Recommended Mechanism; Section 4 list · Checked 2026-10-14
46. 46
    RFC 6263 gives 15 seconds as the minimum recommended keepalive interval Tr for UDP, configurable to larger values.

    Application Mechanism for Keeping Alive the NAT Mappings Associated with RTP / RTP Control Protocol (RTCP) Flows · Section 7 Timing and Transport Considerations · Checked 2026-10-14
47. 47
    RFC 8866 defines a=inactive as a mode in which no media is sent over the stream.

    SDP: Session Description Protocol · Section 6.7.4 inactive (Inactive Mode) · Checked 2026-10-14
48. 48
    RFC 8866 requires an RTP-based system to still send RTCP (if RTCP is used) even when a stream is in inactive mode.

    SDP: Session Description Protocol · Section 6.7.4 inactive (Inactive Mode) · Checked 2026-10-14
49. 49
    RFC 8866 (January 2021) obsoletes RFC 4566 as the SDP specification.

    SDP: Session Description Protocol · Abstract / front matter · Checked 2026-10-14
50. 50
    An inactivity timer that counts only RTP (such as Cisco IOS XE default RTP detection or Asterisk rtp_timeout without rtp_timeout_hold) can fire on a correctly held sendonly or inactive call, because standards allow RTP to stop on hold while only RTCP continues.inferred

    RFC 3264 — An Offer/Answer Model with the Session Description Protocol (SDP) · Section 5.1 and Section 8.4, read with Cisco DevNet media-inactivity-criteria default · Checked 2026-10-14
51. 51
    rtpengine --b2b-url enables an XMLRPC callback when a call is torn down due to packet timeout.

    rtpengine(8) manual page · OPTIONS > --b2b-url · Checked 2026-10-14
52. 52
    rtpengine --final-timeout is a hard cap on call lifetime counted from call creation, and it is disabled by default (0 disables it).

    rtpengine(8) manual page · OPTIONS > --final-timeout · Checked 2026-10-14
53. 53
    Because rtpengine is a media relay controlled by a SIP proxy, an rtpengine media timeout only reaches the SIP endpoints if the proxy or a B2BUA acts on it (for example via the --b2b-url callback); the manual does not describe rtpengine sending a SIP BYE itself.inferred

    rtpengine(8) manual page · OPTIONS > --timeout and --b2b-url read together · Checked 2026-10-14
54. 54
    rtpengine --silent-timeout applies the same dead-stream logic to muted or inactive media streams and defaults to 3600 seconds (one hour).

    rtpengine(8) manual page · OPTIONS > --silent-timeout · Checked 2026-10-14
55. 55
    rtpengine --timeout sets the seconds after which a media stream with no received media traffic is considered dead, defaulting to 60 seconds.

    rtpengine(8) manual page · OPTIONS > --timeout · Checked 2026-10-14
56. 56
    Microsoft says an inbound Direct Routing call that drops 10 to 20 seconds into trying to connect with no audio may be caused by no media information arriving in that timeframe or failed ICE connectivity checks, after which Teams drops the call.

    Issues that affect inbound Direct Routing calls · Calls drop after a specific duration > Calls drop after about 10 to 20 seconds · Checked 2026-10-14
57. 57
    Microsoft attributes Direct Routing calls that drop without an error code after 10 to 60 minutes to session timer or session refresh failures (Session-Expires), not to media inactivity, and advises making the correct refresher send the re-INVITE.

    Issues that affect inbound Direct Routing calls · Calls drop after a specific duration > Calls drop after several minutes · Checked 2026-10-14

Documents

Cite this page

WarmTransfer. (2026, October 14). Calls dropped by RTP inactivity and media timeouts. WarmTransfer. https://warmtransfer.net/knowledge/rtp-inactivity-call-drops

@misc{warmtransfer-rtp-inactivity-call-drops,
  title  = {Calls dropped by RTP inactivity and media timeouts},
  author = {{WarmTransfer}},
  year   = {2026},
  url    = {https://warmtransfer.net/knowledge/rtp-inactivity-call-drops},
  note   = {Verified 2026-10-14}
}