Can I configure the maximum retry times of user-cloudinit-proc-cmdline.service at PXE boot?

[Philip Ye]

Hi,

I was trying my CoreOS's PXE setup for bare metal devices and found something strange: sometimes the bare metal device can be configured with the IP specified in cloud-config-url and public key ingested, but sometimes it just cannot.

Looking into journalctl logs, I found that in the failed cases, user-cloudinit-proc-cmdline.service exited with failure after 15 retries and within every retry, it is complaining for

'network is unreachable'. We have a slow DHCP server and since the DHCP packages are broadcasted across several switches using VLAN, it's expected that there may be some latency for DHCP to finish (possible more than 1min...). So I'm asking if I can change the number of the maximum retry times? Or can I change the maximum sleep time to be larger than the maximum 5s?

Thanks,
Philip

[Alex Crawford]

On 05/02, Philip Ye wrote:
> So I'm asking if I can change the number of the maximum retry times?
> Or can I change the maximum sleep time to be larger than the maximum
> 5s?

Unfortunately, coreos-cloudinit does not let you configure this. This is
something I definately want to add to Ignition [1]. You are now the
second environment to run into this. I'd be willing to accept a PR for
coreos-cloudinit, but we aren't investing a whole lot into that tool
anymore (it's a bit of a dead-end for bare metal).

-Alex

[1]: https://coreos.com/ignition

[Philip Ye]

Thanks Alex. I'm sorry to here that, but may try the PR later if I got time :)