Known Issues and Recommendations
Voice Platform Media Control Platform
The Known Issues and Recommendations section is a cumulative list for all 9.0.x releases of Voice Platform Media Control Platform. This section provides the latest information on known issues and recommendations associated with this product. It includes information on when individual items were found and, if applicable, corrected. The Resolved Issues section for each release describes the corrections and may list additional issues that were corrected without first being documented as Known Issues.
See also Internationalization Issues.
The integration between MCP and SmartHeap is removed in MCP version 9.0.065.95 and later. As a result, the memory usage may be different from older releases (that is, 9.0.065.95 and later) depending on the complexity of VoiceXML applications. Note that the SmartHeap integration is added back to MCP version 9.0.070.43 and later and you’ll not see different memory usage when compared with releases prior to 9.0.065.95.
When using Google Text-to-Speech to play audio prompts to the callers, the audio will not begin to play until the full prompt is synthesized. To enhance the user experience, it is recommended to split the longer prompts into multiple smaller prompts (50 words max in each prompt).
Stereo Opus recordings might fail when the ptime RTP stream is not 20ms.
Workaround: To avoid this, ensure the Session Border Controller (SBC) delivers 20ms ptime RTP streams.
ID: GVP-43688 | Found In: 9.0.048.84 | Fixed In: |
Media Control Platform may terminate if typeahead DTMFs are entered before the <record> tag, in a VXML application.
Workaround: Clear all typeahead DTMFs that are present before the <record> tag.
ID: GVP-22046 | Found In: Unspecified | Fixed In: 8.5.1 |
In rare cases, MCP may take significant time to close the stream with Google Speech-to-Text when an ASR session is finished, in which case a worker thread may be blocked for a while, resulting in ASR errors. You can optionally set vrm.client.grpc.closestreamexplicitly to false to instruct MCP to not close the stream. Errors may be seen on the Google reporting console, but you can ignore those errors.
ID: GVP-43098 | Found In: 9.0.042.31 | Fixed In: |
When the MCP configuration parameter [fm]https_proxy is blank, GoogleASR and GoogleTTS RAP objects have vrm.client.resource.proxy=<http url>, and MCP has "GRPC" in [vrm].client.modules and is connected to Google RAP objects, the Linux MCP Fetching Module uses the GoogleASR and GoogleTTS RAP objects' vrm.client.resource.proxy value when fetching files using HTTPS.
Workaround: Set NO_PROXY=* in MCP startup command line parameters. Example:
NO_PROXY=* LD_LIBRARY_PATH=$LD_LIBRARY_PATH:. ./pwcallmgr -host configserver -port 2020 -app "MCP9041"
The workaround will also disable the use of http proxy (if set in [fm]http_proxy).
ID: GVP-42661 | Found In: 9.0.041.92 | Fixed In: |
Media Control Platform (MCP) v8.5.178.xx does not reuse the SIP request’s TCP connection when responding to the request. Subsequently, MCP tries to initiate a new TCP connection to the FQDN or IP (and port) present in the VIA header of the request.
As a result, MCP might fail to send a SIP response to the request if the remote doesn't accept a new TCP connection. For example, the remote is a F5 load balancer or Genesys Composer.
Workaround: If you have installed MCP v8.5.178.xx, Genesys recommends you uninstall this version and install/upgrade to MCP v8.5.181.78 or a later version.
ID: GVP-35065 | Found In: 8.5.178.xx | Fixed In: 8.5.181.78 |
By default, if the RTCP line is not specified in the SDP offer, MCP assigns the value rtp port +1 to the RTCP port. When the RTCP port is explicitly offered in the SDP message, then the MCP, under very rare conditions, can terminate. To avoid it, just remove the RTCP line from the SDP offer.
Workaround: Remove the RTCP line from the offered SDP.
ID: GVP-24168 | Found In: 8.5.178.78 | Fixed In: 8.5.185.08 |
If the TTS server does not keep the initial transport protocol (that is, does not change from UDP to TCP) within the same session, then MCP might keep using the original transport mode, causing unexpected behavior from the TTS server.
Workaround: Make sure that the TTS server does not change transport protocols during a session.
ID: GVP-24140 | Found In: 8.5.178.78 | Fixed In: 8.5.185.08 |
When the VXML application goes in a loop executing subdialog, the MCP's NGI does not end the VXML application flow on the BYE response from the caller.
ID: GVP-24181 | Found In: 8.5.176.13 | Fixed In: |
Media Control Platform does not upload the unencrypted recorded files (.wav format) to the newly created S3 buckets when the option msml.record.amazonsignedpayload is set to true and when there is a redirection in the POST or PUT requests.
Workaround: Set the options msml.record.updateheader and msml.record.userecordcachedir to true.
ID: GVP-23720 | Found In: 8.5.176.05 | Fixed In: 8.5.178.78 |
In a specific scenario, Media Control Platform (MCP) might rarely generate a memory fragmentation. A memory fragmentation is generated when all of the following conditions are met:
- MCP must perform a very high rate of fetches using TLS mutual authentication.
- The fetched content must not be cached.
- The TCP connections used to fetch the content are not reused, or purely reused, meaning, that for every fetch or couple of fetches a new TLS mutual authentication needs to be performed.
This issue will not be fixed.
ID: GVP-23653 | Found In: 8.5.176.05 | Fixed In: |
For third-party call recording, there are two inbound call legs into the MCP, and each leg negotiates the codec independently with the MCP, and could potentially use a different codec than the other. When starting the recording with a third-party recorder, MCP will try to establish two separate send-only RTP streams with the recorder, regardless of the fact whether a single or dual SIP sessions are used.
When establishing the two recording streams with a third-party recorder, MCP will offer all of the codecs offered by the incoming call legs to the recorder in its INVITE message, and there may be multiple codecs listed in the SDP for each stream. However, each incoming call leg will independently negotiate the codec(s) with MCP and the remote parties will send RTP using the negotiated codec(s). If the recorder, in its answer SDP, does not pick the same codec as the one being streamed by the remote party to be the highest priority codec for each stream, MCP may not send any RTP for the corresponding recording leg; or in some cases, as in the session refresh case below, send RTP using a wrong codec type, albeit without transcoding. This is a defect, related to lack of transcoding support for third-party recording.
Workaround: Set mpc.vrmrecorder.codecpref to "l" (local) in the MCP configuration.
Even after recording has started fine, the recorder could do a session refresh after a while by sending a new offer to the MCP. If the highest priority codec in this offer SDP for each stream is not the one originally negotiated, MCP will start streaming PCMU packets to the recorder, in fact, without actually transcoding the audio from the incoming call leg(s). And this will happen even when PCMU is disabled at the "mpc" level. In such cases, the PCMU or any other unsupported codec can be disabled in the MCP for the recorder by listing only the supported codecs in the [mpc] vrmrecorder.codec configuration option.
ID: GVP-38444 | Found In: 8.5.175.18 | Fixed In: |
When the option msml.record.userecordcachedir is set to false (the default value), the value set in the option msml.record.updateheader is not effective.
Although this scenario is reported as an issue, this will not be fixed since it allows Media Control Platform to update the file headers appropriately.
ID: GVP-23349 | Found In: 8.5.170.71 | Fixed In: |
Media Control Platform might terminate unexpectedly when the Service Quality Analysis (SQA) is enabled, however this is a rare scenario.
ID: GVP-23685 | Found In: 8.5.150.63 | Fixed In: |
Media Control Platform cannot limit the packet size when packetization-mode is set to zero (0), due to a limitation in the library used by MCP for H264 encoding.
This limitation causes issues when using MCP to play video on a Polycom phone using the H264 codec.
ID: GVP-21994 | Found In: 8.5.120.66 | Fixed In: |
You must follow these steps to enable inline grammar access by URL:
Windows IIS Environment
1. Verify that IIS is installed and started, and that MCP is installed.
2. Create and add the directory inlinetmp under C:\Program Files\Common Files\GCTI\www\gvp\mcp\<MCP_application>\grammar\.
The final two steps are automatically performed by the MCP installation in a Windows 32-bit environment:
3. Add the application mcp in IIS, which points to C:\Program Files\Common Files\GCTI\www\gvp\mcp.
4. Enable Directory Browsing and add the MIME type application/octet-stream for the extension * to the application mcp in IIS.
Linux Environment
1. Verify that MCP is installed.
2. Create the virtual directory /var/www/gvp/mcp/ by adding these lines of code to the file /etc/http/conf/httpd.conf:
Alias /mcp/ "/var/www/gvp/mcp/"
<Directory "/var/www/gvp/mcp/">
Options Indexes MultiViews
AllowOverride None
Order allow,deny
Allow from all
ExpiresActive On
ExpiresDefault "now plus 5 minutes"
</Directory>
3. Enable the httpd service
(run chkconfig and specify level 345 httpd on)
4. Start the httpd service
(run start in the directory /etc/init.d/httpd)
5. Add the directory inlinetmp, and grant read and access permission to all files and directories beneath /var/www/gvp/mcp/<MCP_application>/grammar/
ID: GVP-21625 | Found In: 8.5.110.26 | Fixed In: 8.5.120.66 |
Route Unavailable Wakeup functionality does not work correctly when a DNS SRV domain resolves to more than two SRV records.
Workaround: Make sure that the DNS SRV domain resolves to two SRV records.
ID: GVP-21369 | Found In: 8.5.100.85 | Fixed In: |
Media Control Platform may fail to copy a recording file to a drive where MCP is not installed.
ID: GVP-21342 | Found In: 8.5.100.85 | Fixed In: 8.5.120.66 |
Resource Manager may reject a Picture Fast Update INFO request that was sent by Media Control Platform during a bridge transfer.
Workaround:
- Change the method in the bridge transfer to connectwhen = answered.
- Enable PLI (picture loss indicator) requests, instead of SIP INFO based picture fast update requests.
ID: GVP-21415 | Found In: 8.5.050.38 | Fixed In: 8.5.130.82 |
Prompt playback fails when the Media Control Platform logs the ERROR message Invalid Media - Cache track header not valid for media type audio.
ID: GVP-21511 | Found In: 8.5.050.13 | Fixed In: 8.5.120.66 |
The description of the new parameter [fm]revalidatestaleresponse is incorrect inside the MCP configuration XML file (mcp.xml). You can find the correct description in the What's New section of the MCP 8.5.130.93 Release Note.
ID: GVP-22402 | Found In: 8.5.010.30 | Fixed In: |
The first SNMP Get to the Media Control Platform will time out.
ID: GVP-21357 | Found In: 8.1.603.59 | Fixed In: 8.5.120.66 |
Media Control Platform fails to play a stereo MP3 file for a coaching conference scenario.
Workaround: Use a mono MP3 file, and sample at 32KHz maximum.
ID: GVP-21528 | Found In: 8.1.603.57 | Fixed In: 8.5.130.82 |
The take-back of functionality by Media Control Platform at the end of a call fails when the agent disconnects, if MCP is asked to perform a media redirection transfer, when a caller's end supports only audio, but the agent's end of the call supports both audio and video.
ID: GVP-19066 | Found In: 8.1.601.89 | Fixed In: |
CTIC may terminate during a PRACK call flow, if the inbound call uses an offerless INVITE.
ID: GVP-19011 | Found In: 8.1.601.78 | Fixed In: |
Resource Manager does not support key name values that are expressed in UPPER CASE LETTERS in OPM parameters.
ID: GVP-18808 | Found In: 8.1.601.10 | Fixed In: |
You may observe a brief loss of audio at the beginning of a treatment, when you are using CTIC with MCP+NGI to handle treatments.
Workaround: Insert an intentional brief silence at the beginning of the VXML application treatment.
ID: GVP-18039 | Found In: 8.1.502.33 | Fixed In: |
The description of the new parameter [fm]revalidatestaleresponse is incorrect inside the MCP configuration XML file (mcp.xml). You can find the correct description in the What's New section of the MCP 8.5.130.93 Release Note. (GVP-22421)
ID: GVP-22421 | Found In: 8.5.1 | Fixed In: 8.5.140.00 |
MCP may stop unexpectedly when a transfer is terminated immediately after it was started.
ID: GVP-21467 | Found In: 8.5.1 | Fixed In: 8.5.161.34 |
A race condition occurs after Media Control Platform fetches a document, causing the time flag fetchaudiomininum to switch to off when the user disconnects the call.
Workaround: Disable fetch audio, or set fetchaudiominimum to 0.
ID: GVP-20355 | Found In: 8.5.1 | Fixed In: |
When using gvp:dest to write to a file, if "..\" are used to get to the root drive (for example, "C:\ path" or "D:\path"), writing will not work.
That is, if this is the log path:
C:\Program Files\GCTI\VP Media Control Platform 8.1.7\mcp1\logs
Then the following, which should log to C:\ doesn't work:
<log gvp:dest="file:..\..\..\..\..\foolog8.txt"> didnt work </log>
Workaround: Create a hardlinked directory under the MCP's log directory:
mklink /D this_dir_is_a_link c:\install
And then use a <log> tag like:
<log gvp:dest="file:this_dir_is_a_link\foolog8.txt"> ...
ID: GVP-22146 | Found In: 8.5.0 | Fixed In: |
When it is under load, Media Control Platform may throw an error.internal event if the outbound call is declined quickly during a two-call-leg whisper transfer.
ID: GVP-20287 | Found In: 8.1.7 | Fixed In: |
Media Control Platform interprets a caller-inputed DTMF incorrectly—as a different DTMF, but only in the Polish language.
ID: GVP-21297 | Found In: | Fixed In: 8.5.120.66 |
Internationalization Issues
Information in this section is included for international customers. Release numbers in the Found In and Fixed In fields refer to the English (US) release of Genesys Voice Platform unless otherwise noted in the issue description.
There are no internationalization issues for this product.