IP MultiPathing
From: noone (noone_at_noone.org)
Date: 01/26/05
- Next message: John D Groenveld: "Re: Studio 10 compilers for Solaris released"
- Previous message: Richard B. Gilbert: "Re: Solaris 10 ipnodes sourced before hosts"
- Next in thread: Thiemo Nordenholz: "Re: IP MultiPathing"
- Reply: Thiemo Nordenholz: "Re: IP MultiPathing"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Date: Wed, 26 Jan 2005 22:42:30 GMT
Solaris 8
Why does IP MultiPathing require 3 IP addresses to be able to have 2 NIC
fail-over, where one is the active NIC and the other is the standby ?
Seems like a waste of IP addresses.
I know that with IP Multipathing, it pings the gateway and if no ICMP
response arrives back from the gateway, the NIC is set to a failed
status. The problem with this is that, I have experienced a few times
where the colocation switch is actually down .... and thus both NICs are
set to a failed status. However, the internal VLAN is actually still
working. Meaning, on other sparc boxes that did not have IP multipathing
turned on, I can still ssh, ping, etc ... all the other boxes, and I can
also still ping the IP address of host that have IP Multiphating
turned on, though I cannot ssh to it.
Because our database box ( Sun E3500 ) have multipathing turned on, and
all database connections are only internal, from within the VLAN ...
when the switch dies, all the database connections to it are dropped (
and transactions are rolled back ) because the IP Multipathing could not
ping the gateway.
Is there any reason why it does not do something like miitool or ethtool
like in linux, where by you end up with a bond interface and it does not
ping the gateway but it actually checks the link status of the NIC ? In
the scenario above, the NICs are not set to a failed status despite the
gateway being not pingable. Furthermore, you only require 1 IP address
instead of 3.
- Next message: John D Groenveld: "Re: Studio 10 compilers for Solaris released"
- Previous message: Richard B. Gilbert: "Re: Solaris 10 ipnodes sourced before hosts"
- Next in thread: Thiemo Nordenholz: "Re: IP MultiPathing"
- Reply: Thiemo Nordenholz: "Re: IP MultiPathing"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Relevant Pages
|