-
Notifications
You must be signed in to change notification settings - Fork 54
Replace console-based drivers with something more reasonable. #744
Comments
@Izhmash played with SNMP during the summer and concluded that it's not very reliable in setting things. The powerconnect switches have a web interface, which means it has some sort of API. I remember @henn taking about it. But it's undocumented and we would have to "discover" each method by running a web browser in some debug mode. So this option doesn't sound that good either. I don't know what can this be replaced with. |
This doesn't follow; it's entirely possible for it to be just a web UI. And I don't think trying to interact with that programmatically is any better (probably worse). Does the nexus switch also not have any other API? I'm fine closing this if we conclude there's nothing better. |
I might be misunderstanding something then. Isn't the web UI interacting with the switch using some sort of API? In any case, since that is undocumented we can't use it. Let me read the nexus switch's manual to be sure. |
The web ui is running on the switch itself, so whatever mechanism it uses to manipulate the switch doesn't necessarily have to be reachable from the network. |
I was looking at the Nexus Switches we have. We are using our nexus driver against different models.
|
Yeah, our test switch is a lower end model than what's in prod. Cisco gear is expensive. I don't know if a cleaner Api offset the cost of not being able to use our test setup |
Right, we can't replace the driver for the nexus 3548 that we have. We could just write another driver for the other nexus in kaizen (which we have access to) using the API which will probably be more faster and reliable. So I would leave this issue open given that, there's one switch for which we are still using a console based driver that can be replaced with something better. |
Quoting Naved Ansari (2018-01-17 14:01:43)
Right, we can't replace the driver for the nexus 3548 that we have. We
could just write another driver for the other nexus in kaizen (which we
have access to) using the API which will probably be more faster and
reliable.
If we can find a way to test it before deployment I'm OK with this. But
if we can't test it it will probably be less reliable, regardless of the
design.
So I would leave this issue open given that, there's one switch for
which we are still using a console based driver that can be replaced
with something better.
OK with me.
|
The drivers for the dell powerconnect & cisco nexus switches both talk to the interactive console, which was never designed as a programmatic interface. As a result, they're really slow, and particularly in the test suite there's a lot of clumbsy parsing of the output of various commands.
These drivers have their origin in something I wrote way way back, after @pjd-nu had spent about a day fighting with dell's mib files for the powerconnect; we decided it was better to just get something functional, and we'd probably replace it with something more clean and robust later. We still haven't.
The brocade driver is much better -- it talks to an HTTP API that's documented, and actually intended as a programmatic interface. It's also much faster.
When run with the dell driver, the deployment tests take like 5 minutes to run. This came up in conversation while @naved001 and I were waiting for that this afternnoon.
@pns005 suggested that this might be a good project for an intern.
The text was updated successfully, but these errors were encountered: