ICLE failure TOPOLOGY UPDATE

[cash777]

Hi,

i have add a new enclosure (with 10 new blade) to an exisistent - 2 enclosure - CT infrastructure.

After the necessary controls of firmware and di cabling i have initiate
the process to registration of the new enclosure...

now the CT not is plus able to manage 2 enclosure of the previous infrastructure

can i see of previous enclosure only all server as orphan of it (amber color) into component manager

while always into the compoment manager i see the new enclosure and a parent 10 new blade (green color) - but can't managed it

why ct have lost the information data?

i have update the registration date ...Unsuccessfully

note:
ICLE version 1.6;OA fw 2.04;ILO BL460 fw 1.42
separate network management
(ETH0 MA;ILO blade ; OA)

DHCP enabled for ILO , blade ETH0(PXE) and OA

enclosure link enabled

[Robert Crockett · Accepted solution]

Hello Cash777,

The first recommendation is to get your OA's (all of them) up to firmware version 2.12, there were issues with older versions in this regard.

It sounds like you have a networking problem. The ICLE (CT) Server must be able to communicate with the enclosure OA's, iLO2's and the first NIC of each blade server. The default network setup is to allow the ICLE server to assign IP's to the OA's, iLO2's, and the first NIC of each blade server using the ICLE DHCP server.

Do you have enough licenses to register the number of blades you are trying to ?
Have you made any network changes ?
Did you allow the ICLE DHCP server to use it's default scope ? (10.128.0.0 thru 10.128.255.255)

Have you opened up an HP Support Ticket for this issue ? If so, please give me your ticket number so I can see it's status in the system.

Please respond with these answers and we can assist you to get this working properly. :)

best regards,

Robert Crockett

[cash777]

Hi Robert,

thanks for your answer...


HP Support Ticket: 1600171619

I have 49 lincenses which is enough

no network changes have been done and DHCP is working for

OA's, iLO2's and the first NIC of each blade server.

i discovered that the customer has used the old image deployed with

a bl20p using it now for bl460c and the rlx-agent doesn't work

after server reboot

Can it be a cause of the problem?

Is there any other way to re-register servers (which are previously

lost their configuration) without shutting down in order to do PXE-boot?

is RLX-agent able to give information regarding the topology update?



thank you again

Fabrizio

[cash777]

Robert,
only another information...
the customer has configured the bonding (teaming) on two nic host (blade) for the management network ...is this configuration supported?

thanks you again

Fabrizio

[cash777]

Hi Robert,

the ICLE DHCP server does not use the default scope (10.128.0.0 thru 10.128.255.255) but customezed by customer (192.168.2.0 thru 192.168.255.255)

i don't have other information for the time :)

thank you again

Fabrizio

[Robert Crockett]

Hello Fabrizio,
Ok, well using a linux OS image that was built on the p-Class and then deployed to a c-Class is definitely not recommended and I don't know what other issues might arise from that. As for the rlx-agent, it should be installed and activated for each blade. I always recommend running the Provision Task to "Activate" the agent on each blade after a deployed OS is placed on a blade. If the linux OS is installed using ICLE there is checkbox that defaults to being turned on that installs the agent on the target OS and activates it at the same time. When capturing and deploying OS images it is a good practice to re-activate.

The easiest way to tell if the agent is running properly is to run the Provision Task to "Collect blade agent version", if that task returns a version number as it's output... I would believe the agent is working properly. A more thorough test would be to run a Provision task that runs a script and gathers information or installs something. If it accomplishes this with no need to login or enter any kind of credentials, the agent is working properly.
I don't beleive this is the cause of your problem - as the agent is simply a mechanism that allows the ICLE server (linux based OS) to authenticate and log in to the target OS with pre-supplied credentials. This allows ICLE to run scripts and other tasks that need to login (authenticate) to the respective target OS without the need to have a pop-up screen or similar and need human intervention. You can uninstall the agent and you will only lose this functionality, the registration and non-authentication tasks will still work.

The only way to re-register in ICLE is to PXE boot the target node... (sorry) and let the ICLE ramdisk (which is really the HP SmartStart toolkit ramdisk) perform hardware detection and gather the information required to successfully register and populate the ICLE database.

No, per the explanation of the agent above... the agent does not give topology information.

Also, No - NIC teaming is not allowed in our environment on the first NIC... as it breaks our registration process. ICLE must use NIC1 and it's specific MAC address to do it's work and cannot share this NIC (considered to be the "Management" NIC). The ICLE Management NIC can be shared and used for other purposes, although not recommended. One reason is bandwidth, if you share the ICLE Mgmt NIC with say - a database and you provision something that will take a large amount of that NIC's bandwidth like an OS install... there is potential to lose some of the information and corrupt the database. But sharing the NIC with something like a DNS server, etc. should work fine. Also keep in mind that the full height blades have 4 NIC's, so you could successfully team two NIC's as long as neither are the ICLE Mgmt NIC. You can also purchase Mezzanine cards that have 2 or 4 more NIC's on them if two NIC's are not enough.

Hope this gives you a better understanding...

best regards,

Robert

[cash777]

Thanks a lot!

Best regards,

Fabrizio