|
Dear all,
I have set a streaming server on LVS-DR architecture and its setting is like
below.
LVS-DR: one front-end server and multiple real-servers
Front-end server: Linux kernel 2.6.18
Real-servers: Linux kernel 2.6.18 + Darwin streaming server 5.5.5
I use Direct Routing architecture with WRR policy to distribute connections
to each real-server evenly.
However, in my test, if the client watches the streaming movie via LVS, the
streaming connection would be disconnected when 2 minutes period is reached.
This does not occur if the client connects directly to one of the real-servers.
The streaming server uses the RTSP protocol which is based on TCP protocol to
establish streaming connection with clients.
I have tried to use ipvsadm -set and ipvsadm -L -timeout in command line to
check and modify the LVS’ timeout value, and the value of tcp_keepalive_time
(/proc/sys/net/ipv4/tcp_keepalive_time). I also have modified IPVS’ timeout
table in Linux kernel (/net/ipv4/ipvs/ip_vs_proto_tcp.c), but the problem still
remains.
Any suggestions are appreciated.
--==Mailed via NCNU E-Mail==--
_______________________________________________
Please read the documentation before posting - it's available at:
http://www.linuxvirtualserver.org/
LinuxVirtualServer.org mailing list - lvs-users@xxxxxxxxxxxxxxxxxxxxxx
Send requests to lvs-users-request@xxxxxxxxxxxxxxxxxxxxxx
or go to http://lists.graemef.net/mailman/listinfo/lvs-users
|
