Network adaptor not working after failover

Solved5.81K viewsFailover
6
0 Comments

When performing a restart or failover of our DataMiner system, one of the NICs becomes unusable.
Even sending out a ping over that NIC stops working.

Disabling & Enabling the NIC again solves the issue (until the next restart).

Does someone know the exact set of actions that are performed on the NIC during startup/failover?

Ive Herreman [SLC] [DevOps Catalyst] Selected answer as best 6th October 2020

2 Answers

3
1.40K 3 Comments

Hi Ive, When a failover agent goes online, the only operation it does on the NIC level is adding the virtual IP, which is done using windows-level calls. Afterwards an ARP request is sent on the network to notify the network of the new IP. Once the virtual IP is acquired, we will add the SkipAsSource-flag to the primary IP of that NIC. This is done so the server will always use the Virtual IP when communicating with other devices on the network.

The main thing here that can break the NIC would be an IP conflict on the network, specifically when we try to claim an ip which has already been assigned. This can be easily checked in:

  • SLFailover.txt: Added Virtual IP. Context = # / Instance = # (context will be a negative value, and Instance will be 0)
  • Event viewer: event viewer will log this, usually with ID 4199
Ive Herreman [SLC] [DevOps Catalyst] Posted new comment 6th October 2020
Bert Vandenberghe [SLC] [DevOps Enabler] commented

Hi Brent, isn’t there also something being done with the SkipAsSource flag on the IP addresses?
Brent [SLC] commented

Hi Bert, indeed thank you for reminding.
The skipAsSource flag is indeed added to the NIC. With this flag enabled, the Primary IP will be skipped when communicating with other devices on the network, this way the DMA will always respond using the Virtual IP

I added this to the answer
Ive Herreman [SLC] [DevOps Catalyst] commented

Thanks Brent, it seems that we are dealing with a duplicate IP on the network.
0
5.52K 2 Comments

Hi Ive,

Do you have this for all your DMA's?
Could you also share some more information on the Windows version, DataMiner version, driver and HW info on your adaptor?

Michiel Saelen [SLC] [DevOps Enabler] Posted new comment 6th October 2020
Ive Herreman [SLC] [DevOps Catalyst] commented

Hi Michiel,

Thank you for your response.
My cluster contains only one DMA (failover pair).

The system is running DataMiner 10.0.9 on windows server 2012R2.
The NICs are “RedHat Virt IO” network adaptors (which have been working fine for the the last 2 years).
I have 4 NICs but only the primary dataminer nic has this problem.

The logging shows no errors.
Michiel Saelen [SLC] [DevOps Enabler] commented

Could it be related with assigning the virtual IP?