Refresh validation tests

[leec-666]

netlab version 25.11.01
  executable location: /usr/local/bin/netlab
  package location: ['/home/lee/.local/lib/python3.12/site-packages/netsim']
  user defaults: ['/home/lee/.netlab.yml']

Required packages:
  python-box: 7.3.2
  Jinja2: 3.1.6
  PyYAML: 6.0.3
  netaddr: 1.3.0
  ruamel.yaml: 0.18.15
  ansible-core: 2.18.10

Apologies if this is answered in the docs but I've not been able to find the answer: Is it possible to refresh validation tests without restarting the lab?

I'm playing around with validation tests for a lab that takes around a minute to start and it'd be great if there were a way to speed up debugging and testing.

[ipspace]

Apologies if this is answered in the docs

It is not (because it's a pretty shaky "solution")

Is it possible to refresh validation tests without restarting the lab?

netlab create --unlock would disable the "is the lab running" test and recreate the output files. Obviously not a good idea if you changed anything but the validation tests, which is why it's a hidden option.

I could add a "--reload" (or something, do you have a better idea?) option to "netlab validate" to reread and recompile (well, retransform) the topology file, and then use only the validate portion of it.

[leec-666]

netlab validate --reload sounds good. What about an additional flag --reload-and-run that would accept the test names, e.g. netlab validate --reload-and-run ping_again?

[DanPartelly]

Instead of removing the locks on the entire lab, what if we start to treat validation tests completely separately ? Locks on topology are a good thing imo.

The idea is to keep the transformed topology in two caches, one locked for the lab topology and config for the entire lab time and one which can be unlocked and grafted at run-time .

Im not familiar enough with the transformation code to see if this would need extensive changes (then its not really worth it) or can be implemented easily with some well placed indirections.

[jbemmel]

Instead of removing the locks on the entire lab, what if we start to treat validation tests completely separately ? Locks on topology are a good thing imo.

The idea is to keep the transformed topology in two caches, one locked for the lab topology and config for the entire lab time and one which can be unlocked and grafted at run-time .

Im not familiar enough with the transformation code to see if this would need extensive changes (then its not really worth it) or can be implemented easily with some well placed indirections.

Doesn't require separate caches, just a reinterpretation of what "lock" means to Netlab code. Should be straightforward to allow updating the validation tests while leaving the topology as-is

[DanPartelly]

Even better. In this case, we might not even need a --reload flag.

If topology is changed, add warning that topology file is changed, and that only modifications to validation will be considered.

[ipspace]

If topology is changed, add warning that topology file is changed, and that only modifications to validation will be considered.

That's a good idea. Hold my beer 😜

[ipspace]

@leec-666 -- The fix is in PR #2890. If you're running netlab from a local Git repo (see https://netlab.tools/install/clone/), do git fetch --all and switch to the validate-refresh branch with git switch validate-refresh to try out the changes.