noironetworks / opflex / 236 / 1
68%
master: 68%

DEFAULT BRANCH: master
Ran
Files 249
Run time 24s
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: 72.155% (-0.04%) from 72.197%
236.1

push

travis-ci

Tom Flynn 
Loadbalancing stats support for cluster services

- Set DNAT and SNAT flows with cookie and flow_rem flags for stats to get collected
- This is done only for cluster services and only for nexthops that are local to the node to avoid double counting
- Handled transient cases where EP IP can be added/updated/deleted, which could lead to cookie add/delete in SNAT/DNAT flows. To achieve this SNAT/DNAT programming alone has been moved unmodified to programServiceSnatDnatFlows(). The only change here is to install cookie+flag if idgen returns a valid stats cookie.
- idgen.getIdNoAlloc() might return a valid id even if the id was deleted earlier. The deleted ids are stored in erasedId map and later "freed from eraseIds/alloced from Ids" during getId() or garbage collected from Ids map later. Changed getIdNoAlloc() to return invalid if the cookie string is present in eraseIds map.
- during agent restarts, idGen ids will be loaded back to Ids map. SvcTgt flows removal will free up ids for NHs that are valid. This will free up the id only during first time agent run. During restart, freeing up all ids pertaining to the service uuid and it's next hops. Without this, cookie id in SNAT/DNAT flows can be stale.
- Set the priority of * <--> svc-tgt to 97
- make check changes
- verified prom metrics and obs mo stats. Both pod<-->svc and *<-->svc-tgt (based on SNAT/DNAT flows) will increment simultaneously.

Signed-off-by: Gautam Venkataramanan <gautam.chennai@gmail.com>
Change-Id: I7fb4dbd72

17468 of 24209 relevant lines covered (72.15%)

2956.48 hits per line

Source Files on job 236.1