2
39%
master: 39%

Ran 27 Apr 2022 11:46PM UTC

Files 931

Run time 1min

Badge

Embed ▾

Committed 27 Apr 2022 08:57PM UTC coverage: 34.108% (-0.03%) from 34.14%

Job # 41027.2

Build Type

push

travis-ci-com

Committed by

joestringer

Commit Message

Making operator aware of pending pod backlog on nodes for IP allocations

On large clusters with really low pre-allocate values, IP allocations
can be very slow. Especially in scenarios like replacement of a node
with high pod density. This is because everytime maintenance is
performed on a node only pre-allocate number of IPs are allocated in a
batch even though there might be a higher backlog of pending pods
scheduled on the node.

This commit adds a new indexer to the pod informer cache which allows
for querying the cache by node name. This allows the operator to look at
the backlog and allocate more IPs up-to what's available on the
interface. Max above watermark could be used to achieve a similar
effect, but it can result in IP wastage.

Signed-off-by: Hemanth Malla <hemanth.malla@datadoghq.com>

Coverage Stats

47177 of 138315 relevant lines covered (34.11%)

1982.69 hits per line

cilium / cilium / 41027 / 2
39%
master: 39%

README BADGES
x

Markdown

Textile

RDoc

HTML

Rst

Source Files on job 41027.2

cilium / cilium / 41027 / 2 39% master: 39%

README BADGES x

Markdown

Textile

RDoc

HTML

Rst

Source Files on job 41027.2

cilium / cilium / 41027 / 2
39%
master: 39%

README BADGES
x