The tcp_keepalive_interval determines when to recycle dead tcp socket
connections. If you set it too low and a host comes back on that socket
after the socket has been recycled, it may attempt communicating with an
entirely different process, at best triggering failure modes on the apps
communicating over the socket, at worst not triggering failure modes :)
Setting this lower will reduce the overhead of keeping track of would-be
defunct socket connections and can also help application architectures where
there is a high transient socket usage, but it will have zero effect on any
failover situations as it is a passive parameter and not explicitly used to
determine "deadness" of a mpiox path.
-=dave
----- Original Message -----
From: "Vic Engle" <[EMAIL PROTECTED]>
To: <[email protected]>
Sent: Wednesday, May 23, 2007 6:12 AM
Subject: [storage-discuss] tcp_keepalive_interval and iscsi
Hi,
Could anyone comment on any recommended TCP tuning specific to the solaris
iscsi initiator? The tcp_keepalive_interval default seems to be 2 hours.
[EMAIL PROTECTED]>]ndd -get /dev/tcp tcp_keepalive_interval
7200000
Assuming that this parameter is in milliseconds then this works out to
7200 seconds. I believe the sd_io_time default is 60 seconds and can be
retried maybe 3 times. If so then does it make sense to have the keepalive
interval longer than 180 seconds? If I have multiple network connections
to the target and mpxio enabled could I reduce path failover time in the
event that I lose a network path be tuning the keepalive interval down to
60 seconds or shorter?
Thanks,
Vic
This message posted from opensolaris.org
_______________________________________________
storage-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/storage-discuss
_______________________________________________
storage-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/storage-discuss
