Timeout mechanism for Service

[ian.scheidler1]

Hi,

I have a service which configures VPN related data on the devices. In a first test the service went and tried to configure something on the device but that didnt work. It left the device in a sort of loop trying to configure sth. it just couldnt. On the NSO WebUI all I could see was that the green Button wouldnt go to white (old prime UI) and I waited (it didnt go white indicating that no coimmit was done and sth. was somehow stuck).

Then I checked on the device itself, saw it was in a "loop" and cancelled/deleted stuff on the device. In that moment NSO noticed and threw an error which said that a VRF it was in the middle of configuring had vanished (yes, I deleted it on the device to "unblock" it).

Is there an implicit timeout mechanism for services? Do I have to set it explicitly?

I saw in the API there is sth. like an ActionTimeout which can be configured...I also saw there is a data timeout but was unsure for what that would be used.

So, is there sth. similar to ActionTimeout which can be set for a service...or more specifically for all that happens in a cb_create Python method of a service?

[Jan Lindblad]

The problem description is rather vague, so it's hard to understand exactly what's going on here. No commands, no error messages, no code, no logs. If you show more exactly what you are doing and observing, there are many here that should be able to help. What exactly did you see when you saw the device "was in a loop"? The cb_create() is run before NSO even talks to the device so the timeout discussion feels irrelevant here.

[ian.scheidler1]

Hi. Thanks for the reply and sorry for the vague description. It was on site at the customer and I do not currently have access to the logs etc.

You actually partially answered my question though with: "The cb_create() is run before NSO even talks to the device so the timeout discussion feels irrelevant here."

It looks more like the problem occured as NSO was talking to the device, trying to configure sth.

NSO was in the middle of configuring a VRF and wanted to create a RT which already existed. I cannot recall the exact message on the device but nothing in terms of an error message "made its way up to NSO" until we manually deleted the VRF. Then a message popped up in WebUI saying that the VRF NSO was in the middle of configuring has been deleted. But earlier no message about the RT already existing. Maybe we just didnt wait long enough for some sort of timeout mechanism to kick in during the application of the template. Will probably have to test this again and collect logs/error mnessages etc. this time.

Would some timeout mechanism kick in when applying a template/talking/configuring to the device takes unusually long?

[hniska]

The default read and write timeouts towards devices are 20s

admin@ncs% show devices global-settings read-timeout
read-timeout 20;
admin@ncs% show devices global-settings write-timeout
write-timeout 20;