noironetworks / opflex / 315 / 1
68%
master: 68%

DEFAULT BRANCH: master
Ran
Files 253
Run time 17s
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst
coverage: 74.939% (+0.5%) from 74.438%
315.1

push

travis-ci

gautvenk 
Support for nodeport stats with service-pod granularity

- When a LB service is configured in k8s, external ip, cluster ip and nodeport gets allocated. This will also lead to creation of 2 service files by host-agent - 1 external (ext-LB) and 1 non-external (nodeport+cluster). Both will have nodeport set by host-agent.
- A nodePort service will have a cluster IP allocated as well, so we need to support E-W stats for nodePort services as well.
- A cluster service will have just cluster IP allocated; nodeport wont be set.
- Since both cluster and nodeport stats flows are created for the same service file, both the stats are housed under svc-target and svc counter mo.
- Since the non-external service file service-pods (next hops) are a super-set of the external service file's service-pods (only local to the node), node port flows get created only for "non-external" services. Also, this will cover pure node-port services in k8s.
- nodeport stats will get reported whenever there is traffic from "node ip subnet" <--> service-pod.
- optimization: avoided creation of pod<-->svc flows between veth_host_ac and service
- veth host add/delete/modify will lead to updation of these nodeport stats flows.
- Added vethhostac as a new endpoint. Added nodeport stats related tests and other associated tests for veth_host_ac in make check.
- Verified in local environment and tested service-pod add/mod/del; service add/mod/del and vethhostac add/mod/del.
- Will be creating a nodeport service and test further on vagrant setup.

TODO (will be done in followup commit):
Optimization: Need to integrate new service attribute changes (service-type and actual nodeport value (no-op for now)). The stats flows between node <--> service-pod can be avoided if only service-type=cluster services is created.

new flows (veth_host_ac ip is 1.100.201.1/24):
cookie=0x4, duration=79610.812s, table=13, n_packets=0, n_bytes=0, send_flow_rem priority=98,tcp,nw_src=10.0.0.2,nw_dst=1.100.201.0/24,... (continued)

19894 of 26547 relevant lines covered (74.94%)

3389.07 hits per line

Source Files on job 315.1