Last Modified: Oct 06, 2020
See more info
Known Affected Versions:
11.3.0, 11.4.0, 11.4.1
11.5.0, 11.4.1 HF6
Opened: Sep 06, 2013
Related AskF5 Article: K15607
A large data flow back to a BIG-IP system from an ICAP server may be received fully by the BIG-IP system, but not completely sent on to the HTTP server (if REQMOD) or client (if RSPMOD).
The data stream out of the BIG-IP system to the destination server or client stalls and is truncated when the stream returning from the ICAP server to the BIG-IP system completes.
Request adaptation with ICAP returning a large data stream destined for the HTTP server (e.g. a large POST), or response adaptation with ICAP returning a large data stream destined for the HTTP client (e.g. a video stream); and HTTP server or client is slower than ICAP server sending the data, such that flow-control is activated.
Only a partial workaround is known, and might not be possible in all cases. If possible, ensure that the destination server or client connection is at least as fast as the ICAP return connection, or conversely ensure that the ICAP return is slower, so that flow-control is not generally needed.
Data flow returning from the ICAP server to the BIG-IP is correctly flow-controlled to match the rate at which the destination server or client can consume it, and the complete data stream reaches the destination.