Very late ILO access

[mururoa]

Hi there,
We have a problem since years that really is annoying now.
The problem is that when we setup correctly an ILO card, in servers or blades, it takes HOURS for the card to be reached.
Just yesterday after a motherboard change on a blade and after ILO configuration it took about 2h30 for the ILO card beeing reachable from external computers as well as from the OA !
In the mean time no ping or nothing else works and suddenly without doing anything we can connect to the address configured.
Same if I unplug the ILO cable from one server and plug it on another server with same ILO configuration. It take hours before I can reach the new server.
I cant see any parameter on the ILO configuration that can explain that.
Ofc it's more that annoying after a MB change that you cant access the server with ILO as well as with network interfaces since MAC address has changed and linux dont recognise his old eth0 and eth1.
Any idea ?

[Terry Hutchings]

I haven't seent his problem before. Are you connecting to the ILO with a browser? If so, which one? Also are you using the IP address or DNS name for the ILO? Does the same type of behavior happen on all servers or just your blades?

[Kase]

I haven't seen an issue like this before, either. If this issue happens for all servers, I would suspect something with the network is at fault. Is the error condition described true for all servers? Approximately how many? I would start troubleshooting by:

1. Swap iLO cables between two servers to create the error condition. Verify the error exists by checking IP connectivity to both servers' iLO.
2. Unplug one test server's iLO port and plug directly into that server's iLO port via crossover cable with a laptop (or another server). Check IP connectivity. If you can connect via crossover cable as described, you have probably isolated it to a network infrastructure issue.

[mururoa]

@Terry : any browser AND from the OA. Using IP @. All servers.

@G : 1 - done and this take hours so that each server can be reached using the 'new' cable.
2 - hard to setup since I have no laptop but the idea is good. I have to test that.

[redfacebob]

use wireshark on the management network, examine the conversation.

Are there errors/retries reported by the nic. In linux, look at ipconfig output.

[Adrian Clint]

PS: To connect to a blade OA you do not need a crossover cable.
A straight one works as well.
Just make sure you plug into the active one.

[mururoa]

I've done the direct link test and the result is instant connectivity.
So the problem should be on network equipement configuration like routers and switchs on the administration network.
What parameters should I look at in these devices ?

[Kase]

As redfacebob pointed out earlier, to troubleshoot the network issue you would need to set up wireshark and examine the conversation between a management station and a server with iLO exhibiting the error. At this point, it sounds like a switch configuration issue.