|
What kernel are you using? Are you also using the FTP helper module?
2.6.17.8 vanilla kernel
no ftp involved..
client(rdp client) -tcp-> dr-lvs -tcp-> real server(rdp session)
<------return tcp---------
dr-lvs has 1 active record from cli -> server
--
then the client closes his/her rdp client with out logging off of the rdp
session
client() X dr-lvs -tcp-> real server(rdp session)
dr-lvs has 1 inactive record from cli -> server
--
so far so good if the client reconnects in under 10minutes they will get
their old session back yay!
however after 10min~ (persistance is set to 1day[86400seconds]) we get
client() X dr-lvs X real server(rdp session)
dr-lvs has no records (unhappiness)
if a client then connects to the service they are loadballanced as per
expected for a new session. if they land on the realserver with their rdp
session on it, yay they are reconnected and happy. If they land on a
realserver that doesnt have their session on it.. then a new one is
created and the old one is also running on the other real server = sadness
--
service is setup like...
-A -t server:3389 -s wlc -p 86400
after reading all avilble documentation i was under the impression that
this would then keep a persistance record for cli -> server up for a day.
Weather the original was open (active) or if it was in a closed (inactive)
state(no current connection). However data is being flushed in 10~minutes.
Is this due to the tcp_keepalive_time as Joseph Mack has pointed out?
(from the kernal docs it looks as if a 11~ minute timeout happens on
inactive tcp connections), is this a ipvs issue (ie even though a
persistance of 86400 is set, kernel closes out the record anyway maybe due
to the keepalive)?
-joseph
_________________________________________________________________________
Info: Email:
Joseph T. Duncan duncan@xxxxxxxxxxxxx
109 Kidder Hall
Corvalis Or 97333
| 