Hi.
I have a child-proxy which gives it's users timout-errors, while it is getting
data from its parent. Mainly this occurs while downloading big ftp-files
(>1MB)
The configuration:
Parent-Cache <--64k/s--> Child-Cache
1.NOVM.2[02] 1.NOVM.22 or 1.1.6
The Child-Cache requests a ftp-object which is already in the cache of the
parent.
It asks per ICP for it an gets (as expected) an UDP_HIT
907162812.057 10 194.233.113.242 UDP_HIT/000 58 ICP_QUERY
ftp://ftp.mindjet.com/pub/mm30-3g.exe - NONE/- -
Then it starts to download the object... I can see the transfer on my
networkmonitor. It runs at maximum rate (all available bandwith on the
64k/s-line).
But (and that is the problem) the requesting user sees nothing, his browser
does not show that there is data approaching... after 900 seconds it gives a
READ TIMEOUT to the user.
Child: 907163978.150 900242 172.21.65.6 ERR_READ_TIMEOUT/400 801 GET
ftp://ftp.mindjet.com/pub/mm30-3g.exe - PARENT_HIT/proxy.hx.abacom.net
-
Parent: 907163905.430 1092862 194.233.113.242 TCP_HIT/200 4522406 GET
ftp://ftp.mindjet.com/pub/mm30-3g.exe - NONE/-
application/octet-stream
(ok, the clock of the Child was 5min out of sync)
on smaller objects the Browser also stalls. After _completing_ the transfer
between the caches it delivers the data as fast as it can to the client.
An upgrade from 1.1.6 to 1.NOVM.22 does not change the behaviour.
If I let the child go directly, there are no problems... it loads the data
from the source and delivers the data simultanously to the client.
How can i get this behaviour also with a proxy-cascade? (other proxy-chains
here work without problems...)
Both proxies are configured to the installation-defaults.
Hints?
Cord
-- Systemadministration Email admin@abacom.net Abacom-Internet GmbH & Co WWW http://www.abacom.net/ Konrad-Zuse-Str. 1 Tel 05271/967-0 37671 Hoexter Fax 05271/967-267Received on Tue Oct 13 1998 - 05:39:21 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:42:28 MST