varnishadm fails on long responses from varnish (patch attached)

Martin Blix Grydeland martin at
Thu May 26 17:57:15 CEST 2011


I've found that varnishadm will sometimes exit thinking it has lost the CLI
connection to varnish when the result is long (exceeding the TCP MTU I
guess). This is because the socket it is listening on is non-blocking, and
the read_tmo() function will not restart a read operation that failed with
EAGAIN. Patch attached will change this function into restarting the poll
with timeout on EAGAIN.

Please review - any comments appreciated.

Martin Blix Grydeland

Martin Blix Grydeland
Varnish Software AS
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0001-Allow-restart-of-timeout-in-read_tmo-when-doing-non-.patch
Type: application/octet-stream
Size: 1009 bytes
Desc: not available
URL: <>

More information about the varnish-dev mailing list