Closed raddessi closed 6 years ago
raddessi
commented
6 years ago Similarly, I just ran in to the issue that passing interfaces.interface.Ethernet1.config.enabled = False will successfully disable an interface by setting "shutdown" on the EOS switch I'm testing on, but then flipping it back to True won't change anything (by passing a no shut) so it just continually fails the compliance report afterwards.
dbarrosop
commented
6 years ago All that is supported and should work, you are probably not merging properly the candidate into the running configuration. Would you mind sharing your code so I can verify my assumption?
dbarrosop
commented
6 years ago Ok, I did some testing and you found a bug in the mappings. I suspect your code wasn't correct anyway so, please, share it :P Will try to push a fix soon.
dbarrosop
commented
6 years ago So this was a combination of assumptions and a bug you uncovered. I have fixed it in #135. Would you care to test?
To check the current behavior, running the code:
from napalm_yang import base, models
running = base.Root()
running.add_model(models.openconfig_interfaces)
running.add_model(models.openconfig_interfaces)
running.interfaces.interface.add("et1")
running.interfaces.interface["et1"].config.enabled = False
running.interfaces.interface["et1"].config.description = "asdasd"
candidate = base.Root()
candidate.add_model(models.openconfig_interfaces)
candidate.interfaces.interface.add("et1")
candidate.interfaces.interface["et1"].config.mtu = 9000
candidate.interfaces.interface["et1"].config.enabled = True
print("running")
print(running.translate_config(profile=["eos"]))
print("translate")
print(candidate.translate_config(profile=["eos"]))
print("merge")
print(candidate.translate_config(profile=["eos"], merge=running))Should print:
running
interface et1
description asdasd
shutdown
exit
translate
interface et1
mtu 9000
exit
merge
interface et1
mtu 9000
default description
no shutdown
exitThe following should also work as you expect:
candidate = base.Root()
candidate.add_model(models.openconfig_interfaces)
candidate.interfaces.interface.add("et1")
candidate.interfaces.interface["et1"].config.description = ""
candidate.interfaces.interface["et1"].config.mtu = 9000
candidate.interfaces.interface["et1"].config.enabled = True
print("translate")
print(candidate.translate_config(profile=["eos"]))Prints:
translate
interface et1
mtu 9000
default description
exitOk, just noticed there seems to be an issue with enabled still. I think it's a bug on pyangbind, will dig into it.
dbarrosop
commented
6 years ago For reference, this would fix that problem:
https://github.com/robshakir/pyangbind/pull/203
Let's see what they say.
raddessi
commented
6 years ago Hey you are fast :) I'm actually using napalm-salt to generate the config via the netyang module, I can share any data you would like but I haven't dug deep enough in to the netyang module to figure out where to dump the raw yang config yet.
Below is output from when I have manually logged on the switch and shut down a port on it, where the compliance report has failed due to it not bringing the port back up.
Here is the pillar I'm passing in to netyang.managed:
openconfig_interfaces:
----------
_kwargs:
----------
filter:
True
interfaces:
----------
interface:
----------
Ethernet1:
----------
config:
----------
description:
no description
enabled:
True
mtu:
9000
Ethernet2:
----------
config:
----------
description:
no description
enabled:
True
mtu:
9000
And here is the response from the state run:
----------
ID: interfaces_config
Function: napalm_yang.managed
Result: True
Comment: Already configured.
Loaded config:
interface Ethernet1
mtu 9000
description no description
exit
interface Ethernet2
mtu 9000
description no description
exit
Started: 12:28:36.546312
Duration: 3858.276 ms
Changes:
----------
compliance_report:
----------
complies:
False
skipped:
to_dict:
----------
complies:
False
extra:
missing:
present:
----------
interfaces:
----------
complies:
False
diff:
----------
complies:
False
extra:
missing:
present:
----------
interface:
----------
complies:
False
diff:
----------
complies:
False
extra:
missing:
present:
----------
Ethernet1:
----------
complies:
False
diff:
----------
complies:
False
extra:
missing:
present:
----------
config:
----------
complies:
False
diff:
----------
complies:
False
extra:
missing:
present:
----------
description:
----------
complies:
True
nested:
False
enabled:
----------
actual_value:
False
complies:
False
expected_value:
True
nested:
False
mtu:
----------
complies:
True
nested:
False
nested:
True
nested:
True
Ethernet2:
----------
complies:
True
nested:
True
nested:
True
nested:
True
diff:
loaded_config:
interface Ethernet1
mtu 9000
description no description
exit
interface Ethernet2
mtu 9000
description no description
exit
Let me see if I can get it to print the raw yang it would generate from the pillar...
I can test https://github.com/napalm-automation/napalm-yang/pull/135 if you would like still.
raddessi
commented
6 years ago Oh, yeah, https://github.com/robshakir/pyangbind/pull/203 totally seems like the right way to go.
dbarrosop
commented
6 years ago It's fine, originally I made the assumption that "default" values would be ignored unless there was a "running" config indicating otherwise (note that in my example above "merge" was as you probably expected). That assumption is probably going to lead to confused users so let's go with robshakir/pyangbind#203 if they agree.
raddessi
commented
6 years ago Awesome, looks like they merged that in now. Do you need to do anything else here or should I test?
dbarrosop
commented
6 years ago Yeah, I need to regenerate the bindings with new pyangbind. Give me a couple of days to do so and then I will give you a branch so you can test.
dbarrosop
commented
6 years ago did you have the chance to test #135?
dbarrosop
commented
6 years ago Closing, let me know if the mentioned PR doesn't solve the problem.
I may be dense but I can't seem to find any documentation about how to perform a
no <something>. Specifically in this case I'm just trying to wipe the description on a port. I've tried setting it to False/None and an empty string, False/None just get converted to strings and an empty string fails entirely with the cli complaining about an incomplete command.I am running with:
I can think of a few other cases we would need to
no <something>though, is this a supported feature?