quick_abort again

 Hello All !
The question I want to ask is, probably, number one in squid users top of
faq. But, unfortunately, that thing is not clear to me.
I have Squid 1.NOVM.22 (builded from source, not rpm binaries) on Linux
2.0.35 (RH5.1)+filehandle-v8.01 patch. Everithing works fine. But
squid immediately stops retrieving the object after user has pressed
<cancel> button in browser window (or connection between squid and browser
fails). In the cache.log message ERR_CLIENT_ABORT appears. I had checked
squid.conf and found line
quick_abort -1 0 0
that line disables quick_abort feature, as I know from the squid docs and
this mailing-list archives. Strange ...
I put in squid.conf
quick_abort 10000 10 20000
then type
squid -k reconfigure
I'm sure Squid really cought up SIGHUP and really reread his squid.conf
and start repeat tests ...
While browser is downloading files, squid stores it to disk, objects exist
in VM Objects and Objects queries as CacheManager says. I wait until 1000k
of 2000k file downloaded ( 1000<10000 and 1000 not > 20000 and 50%>10%) and
cancel download. I think squid will continue download, but what a hell !
Squid aborts retrieving, file removed from VM Object and Objects, I dont
believe it and after some minutes trying repeat download the same file from
the same URL and I'v got TCP_MISS message in access.log (of course ...).
Very strange ...
And I shift:
quick_abort 0 10 20000
quick_abort 20000 10 20000
quick_abort 10000 50 20000
and so on, many, many times -- the result is the same -- Squid NEVER will
continue the object retrieving !
PLEASE HELP, WHATS WRONG WITH ME ? My head is sick or Squid always has no
ability finish such operation ?
Thanks for reading.
                                                                      

                Victor Y. Kovalenko, Moscow, Russia
P.S.
Sorry for bad english.

                                                                      
Received on Thu Sep 17 1998 - 01:44:50 MDT