Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/jbemmel/srl-opergroup-agent
https://github.com/jbemmel/srl-opergroup-agent
Last synced: about 1 month ago
JSON representation
- Host: GitHub
- URL: https://github.com/jbemmel/srl-opergroup-agent
- Owner: jbemmel
- License: apache-2.0
- Created: 2021-07-19T19:29:01.000Z (over 3 years ago)
- Default Branch: main
- Last Pushed: 2022-03-08T00:43:36.000Z (almost 3 years ago)
- Last Synced: 2024-10-05T17:21:49.201Z (4 months ago)
- Language: Python
- Size: 354 KB
- Stars: 2
- Watchers: 3
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# srl-opergroup-agent
SR Linux agent to implement operational groups where target elements are administratively disabled if a monitored link or session goes down
# Use case: uplink monitoring
Monitor operational state of a link:
```
gnmic -a clab-opergroup-lab2-spine1:57400 -u admin -p admin --skip-verify -e json_ietf get --path /interface[name=ethernet-1/1]/oper-state
```
The result will be "up" or "down". Accordingly, set admin-state to "enable" or "disable" for all target elements:
```
gnmic -a clab-opergroup-lab2-spine1:57400 -u admin -p admin --skip-verify -e json_ietf set --update-path /interface[name=ethernet-1/2]/admin-state --update-value "disable"
```
Similarly, a BGP session can be monitored or disabled. Generator expressions can be used to enumerate multiple targets (expanded to multiple gNMI SET commands):
/interface[name=ethernet-1/(2|4|5-7)]/admin-state
# Demo scenario
Inspect the uplinks and host facing links:
```
info from state interface ethernet-1/{1..4} | filter fields admin-state oper-state
```
Check the oper-group policy:
```
info /opergroup-agent
```
Take down first uplink:
```
enter candidate
set /interface ethernet-1/1 admin-state disable
commit stay
```
Check that opergroup agent brought down the access ports:
```
info /interface ethernet-1/{3,4}
```
# Demo scenario 2: BFD sessions
On Spine1, show BFD enabled interfaces (OSPF on links and BGP EVPN on loopback):
```
info from state /bfd | filter subinterface
```
On Leaf1, currently all BFD sessions up:
```
info from state /bfd network-instance default | more
```
Bring down BFD on the Spine1 loopback:
```
enter candidate
set /bfd subinterface lo0.0 admin-state disable
commit stay
```
Check the agent state on leaf1:
```
info from state /opergroup-agent
```
Also bring down OSPF on Spine1, which ends the 2nd BFD session:
```
enter candidate
set /network-instance default protocols ospf instance main admin-state disable
commit stay
```
Check the agent state on leaf1 again:
```
info from state /opergroup-agent
show /interface ethernet-1/3
```
## Caveats and learnings
Currently (SRL 21.6.2) no gNMI 'delete' events are generated for /bfd/network-instance[name=default]/peer[local-discriminator=*]/oper-state, and sessions that go down
simply disappear after a while. This could cause the agent to retain stale state for BFD sessions that no longer exist, counting them as 'down'.
To resolve this, the agent could retrieve the updated state for all entries when triggered by an on-change event - but this still does not 'remember' downed sessions.
A stable trigger condition is the desired total count of BFD sessions