this could seem odd but it is not.
I have a complex environment composed mostly by p570 POWER6 machines.
Each machine has 2 VIOS (last update of 1.5) that gives vSCSI and vETH
connectivity to LPARs. VIOS are configured to guarantee redundancy
(control channel for SEA and dual vSCSI path) The environment hosts
several customers, each with two nodes in cluster on different
machines.
Whenever a new customer ask for a new couple of LPARs the datacenter
guys delivers the new LPARs. Each couple has a couple of VLAN IDs for
management and service, ant their virtual ethernet adapter are
configured like:
Virtual ethernet adapter
Adapter ID :
2
VLAN ID :
88
Required :
False
IEEE 802.1q compatible :
False
Additional VLAN IDs :
Access external network :
False
Trunk priority :
0
Now the question is:
Having SEA adapters built like this (this is a sample of the "main"
adapter, obviously VIOS network redundancy is guaranteed through
control channel):
Virtual ethernet adapter
Adapter ID :
11
VLAN ID :
99
Required :
False
IEEE 802.1q compatible :
True
Additional VLAN IDs :
11 22 33 44 55 66 77 88 (and so on...)
Access external network :
True
Trunk priority :
1
Is there a way to update "Additional VLAN IDs" parameter without
rebooting VIOS????
OK the answer could be "having VIOS redundancy, systems will not be
affected by VIOS rebooting one at a time" but we have some problems in
the environment that prevents us from rebooting VIOS freely...(don't
want to give too many details here...) and we can't rely on clustering
system implemented...so we need to know if there is a way to insert
new VLAN IDs in "Additional VLAN IDs" parameter not having to reboot
VIOS each time a new customer arrives...
(I hope I've been clear enough :-S )
Any help will be appreciated.
Thanks,
Thomas
I second that. Removing and readding the SEA or rebooting each VIOS
are the only ways we found we're able to change the additional VLANs
option. Yes, this is pretty inconvenient and i wonder if there's
really no way at all that IBM could have implemented this as a DLPAR
or other dynamic configuration operation(?)
If you choose the DLPAR way, make sure the configuration changes
find their way in the partition profiles of the VIOS, to avoid any
ugly surprises ;-)
Best regards,
Frank
Ok,
everything seems to work.
I followed your suggestions, but came out that was impossible to
remove the old adapter and add the updated one due to an error
HSCL151A "A virtual Ethernet trunk adapter with the same MAC address
already exists", even if I removed every reference to the old
adapter...
So I tried another way, and this is the whole story:
- ha_mode=standby on the second VIO
- removal of the sea adapter, as well as the two virtual adpater used
(primary and ctl_chan) and removal thorugh DLPAR operation
- creation of a new/updated primary adapter (old VLANs + new VLANs) -
trunk piority 2 --> this gave the 151A problem
- creation of the old adapter with old VLAN IDs
- creation of another adapter with new VLAN IDs - trunk prio 2 - this
is compulsory
- from the second VIO $ mkvdev -vdev $ENT_OLD,$ENT_NEW etc etc with
ha_mode=auto
- from the first VIO # chdev -l $SEA_OLD -a virt_adapters=$ENT_OLD,
$ENT_NEW
- this came up with first adapter working perfectly and the second one
(with new VLAN IDs) in Active=False on both VIOs
- on the first VIO (the one handling network connections)
ha_mode=standby --> this led the second VIO to activate the new
adapter
- on the first VIO ha_mode=auto
After double-checking that everything was allright I saved current
configuration of course ;-)
So, thank you very much for your support, I hope that this could be
helpful for the future.
Thomas
the answer from IBM is in this document : http://www.redbooks.ibm.com/redpapers/pdfs/redp4194.pdf
under the topic of "Changes to the VLAN".
After DLPAR a new trunk vETH you should do a "chdev -dev ent4 -attr
virt_adapters=ent5,ent8" on both VIOS if the new vETH that was added
was ent8, and the vETH you currently were using is ent5.
Regards.
1) On the VIO server DO NOT add any vlans to the existing virtual
adapter, instead, you will create a NEW virtual ethernet adapter, give
it a vlan id that is not being used, nor will ever be used. Then give
it an "additional vlan id" of the actual vlan you want to use for vlan
tagging.
2) Next you will shutdown your vioserver, then reactivate the
vioserver (if you did not use Dlpar)
3) At this point you should have your VIO server up and running with a
new ent<number> for the new virtual ethernet adapter that you created.
4) Now you will add this new virtual ethernet adapter to the SEA (yes,
now you will have 2 virtual adapters in your SEA).
$chdev -dev entX -attr virt_adapters=entY,entZ (where entX is the SEA
and entY is the adapter that was already part of the SEA and entZ is
the new virtual adapter.
Remember to use the same "Trunk Priority" on the new trunk adapters as
the ones in the SEA that your about to use.
/Jesper
"To start bridging a new tagged VLAN ID, you can create a new virtual
Ethernet with a
throwaway PVID and any tagged VIDs you want. Then, use dynamic LPAR to
move it into the
Virtual I/O Server, and use the chdev command to add the new adapter
to the list of
virt_adapters of the SEA. It will immediately begin bridging the new
VLAN ID without an
interruption."
OK, so I've got a trunk setup on my VIO server with a non-used primary
vlan id, and several tagged vids. These work on the VIO server, so
I'm assuming I've done that bit right! Do I need to have a virtual
interface on the VIO server for every vlan I want to use on the other
LPARs in my system?
The section in the redbook then goes on to talk about moving from
untagged to tagged virtual ethernet adapters but it's not clear
whether this is talking about the VIO server or the other LPARs, or
indeed why you would want to do this!
Is there any clearer documentation on this anywhere as I find the IBM
docs almost incomprehensible?! :(