Unable to reconnect namespace via NVMe/TCP

[Maurizio Lombardi]

On Mon Aug 18, 2025 at 10:58 PM CEST, Chris Leech wrote:
> On Tue, Aug 12, 2025 at 06:48:29PM +0300, Anton Gavriliuk wrote:
>> Hi
>> 
>> There are NVMe/TCP target and initiator servers, both running on
>> RHEL10 (6.12.0-55.25.1.el10_0.x86_64)
>> 
>> NVMe/TCP target exports single NVMe SSD
>> 
>>       "namespaces": [
>>         {
>>           "device": {
>>             "nguid": "01000000-0000-0000-8ce3-8ee3064aa4f2",
>>             "path": "/dev/nvme0n1"
>>           },
>>           "enable": 1,
>>           "nsid": 1
>>         }
>>       ],
>> 
>> If NVMe/TCP target is not available, initiator tries to reconnect
>> every 10 seconds
>
> How is the target becoming unavailable?  Is it as network interruption,
> or is the target being rebooted or reconfigured?
>
> Is the shared snippet the entire "namespaces" section of the
> configuration file?
>
> It looks like the target code will generate a random uuid for a device
> when it's configured, which could then trip up the host attempting to
> reconnect across a target reboot. But I think the uuid can be saved as
> well as the nguid in the target configuration.

Indeed, normally, nvmetcli saves the uuid in the config file.

Maurizio