Ken Aaker wrote:
It seems to me there is possibly some confusion of terminology here...
There should be no way that the iSCSI target drives themselves should
be listed in the VLDB. If additional ethernet cards were added to the
server to connect to the iSCSI targets, they would be treated the same
as any other ethernet interface, since they're just carrying TCP/IP
traffic. If clients have viable routes to the added ethernet
interfaces in the server, the openafs traffic might reduce the
bandwidth available to the iSCSI connections, but other than that it
should work.
We are using routable addresses for iSCSI target disks, although I
thought those addresses were being firewall restricted from user/client
subnets. So I guess they do qualify as viable routes. Yes everything
works properly, but some users are concerned, as you suggest, that
performance will be affected if clients keep trying to use those addresses.
I have added a NetRestrict file to a newly built fileserver and in fact
the iSCSI addresses are not showing as listed in the VLDB. My problem
remains to remove existing addresses. Jeffery's solution, delentry and
syncvldb, seems faster than moving volumes. I'll set up the schedule
for an off peak window and give it a go.
If the clients don't have viable routes to the additional ethernet
interfaces on the server, but are still trying to use those addresses,
there's something else going on....
I don't think anything unseen is happening here. I think it's a simple
case of me not restricting the usable address space, initially.
Ken Aaker
_______________________________________________
OpenAFS-info mailing list
[email protected]
https://lists.openafs.org/mailman/listinfo/openafs-info
--
veritatas simplex oratio est
-Seneca
Andrew Bacchi
Systems Programmer
Rensselaer Polytechnic Institute
phone: 518.276.6415 fax: 518.276.2809
http://www.rpi.edu/~bacchi/
_______________________________________________
OpenAFS-info mailing list
[email protected]
https://lists.openafs.org/mailman/listinfo/openafs-info
