Bug 200290321: NetworkManager-wait-online.service disabled, why?

Autonomous Machines Jetson & Embedded Systems Jetson Xavier NX
#1

Platform: NVidia NX w/ L4T 32.4.4 w/ custom Ubuntu 18.04 server based rootfs

I was hoping to get some clarification on why NetworkManager-wait-online.service is disabled in the nv_tools/scripts/nv_customize_rootfs.sh script (in L4T 32.4.4)? It mentions (I assume an internal) bug number 200290321.

We have an interest in using the NetworkManager-wait-online.service to gate the network-online.target on system boot.

We are also seeing weird behavior with the following kernel messages appearing several minutes after boot:

Mar 10 15:59:37 ws-nxcore kernel: random: crng init done
Mar 10 15:59:37 ws-nxcore kernel: random: 7 urandom warning(s) missed due to ratelimiting

and don’t know if it is somehow related.

The reason I am drawing the relation is because of the following behavior:

With NetworkManager-wait-online.service enabled the pollinate service times out after 5 minutes during system boot, blocking ssh from starting.

  1. The reason this occurs is because the pollinate service waits for network-online.target.
  2. With the NetworkManager-wait-online.service enabled, the pollinate service starts after there is a confirmed network online because the network-online.target isn’t triggered until the NetworkManager-wait-online.service confirms there is an online network
  3. For some reason the pollinate service takes ~5 minutes to run and eventually times out (being unsuccessful), allowing the ssh service to start. (the pollinate service puts a restriction that it must run before ssh)
  4. Around this time (but I believe unrelated) the above mentioned crng init done logs appear.

With NetworkManager-wait-online.service disabled the pollinate service runs when there is no online network, exits quickly (with error I believe), allowing ssh to start quickly.

  1. Since the network-online.target is not gated by anything it finishes right away (before the network is confirmed to be online)
  2. Therefore pollinate runs right away (when there is no network) and exits right away as the //entropy.ubuntu.com/ is not reachable
  3. ssh is also unblocked because pollinate finished quickly (again, with error).
  4. Several minutes later the above mentioned crng init done logs appear

Neither of the above 2 scenarios is ideal.

Is the above mentioned bug (200290321) related to the late arriving crng init done message? Is there a way for me to resolve my issue to have NetworkManager-wait-online.service enabled and the pollinate service not hang?

Thanks,
Joe

#3

Hello,

This service has been disabled for almost 4 years.

The reason we disabled this service is:

This service wait until network manager becomes online. This service is really not necessary because most network
daemons start properly even if network has not been configured yet.

And since this service has the timeout specified to 30s it will fail most of the times as network may take time to configure. This results in stalling some of the services and resulting in slower boot.

Thus, disabling this service.

If you need such service, just enable it by yourself. You can just remove those lines in apply_binaries.sh

#5