Hi James,
James Carlson wrote:
Jack writes:
We're working on the iSCSI initiator in Solaris, it utilizes TCP/IP
protocol in kernel (so*) to enumerate block devices on the local host.
Users may need to access these block devices in an early booting up
time, e.g., by mount points in /etc/vfstab, or use them as shared DID or
quorum device in Cluster environment.
The problem sounds similar to booting via NFS. I'd suggest that you
work with the NFS folks to make sure that the same solution that works
there works for boot via iSCSI. (Robert Thurlow is probably a good
contact.)
Thanks for the into, I'll try to have some clues from there.
In general, we rely on the booter to set up whatever routes it needs,
particularly so if it must run before normal system services have
started running.
I'm curious about if there is a service which is configurable, say,
online until a specified IP is routable. If so we can add a dependency
on it.
No, that's just impossible. It's in the nature of TCP/IP.
Other than simply sending packets to a given IP address and waiting
for a reply, there's no way to know whether the routes in the kernel's
forwarding table are sufficient to reach that given destination.
Using RTM_GET on a routing socket will tell you whether a route to a
destination exists, and you can listen to routes as they're being
installed and removed, but it won't tell you whether any matched route
is really the right one.
Even if you do have the right one, it's impossible to tell whether any
of the intermediate nodes will forward your packet for you -- without
sending the packet.
OK, here I have the picture, thanks. Originally we're expecting a
milestone to indicate the routing service is full functional, now seems
it is still not a complete solution.
Best regards,
Jack
I think the best answer is the simple one: send and retransmit as
necessary.
_______________________________________________
networking-discuss mailing list
[email protected]
