[engine-1.21] Reduce node startup time

[brandond]

Proposed Changes

• If using the built-in CCM, wait for it to clear the uninitialized taint before starting the scheduler to ensure that pods can be scheduled immediately.
• Optimize agent startup time by watching the local node instead of getting/sleeping in a loop.

Types of Changes

bugfix

Verification

See linked issue

Linked Issues

• [release-1.21] K3s server startup time greatly increased #4360

User-Facing Change

Resolved an issue that would occasionally cause K3s to take more than a minute to schedule initial pods on the first server in a cluster.

Further Comments

[codecov-commenter]

Codecov Report

Merging #4362 (2ad5925) into engine-1.21 (998e2bb) will decrease coverage by 0.03%.
The diff coverage is 1.66%.

@@               Coverage Diff               @@
##           engine-1.21    #4362      +/-   ##
===============================================
- Coverage        11.47%   11.43%   -0.04%     
===============================================
  Files              139      139              
  Lines             9036     9067      +31     
===============================================
  Hits              1037     1037              
- Misses            7774     7805      +31     
  Partials           225      225              

Flag	Coverage Δ
inttests	0.67% <1.66%> (-0.01%)	⬇️
unittests	11.09% <1.66%> (-0.04%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
pkg/agent/flannel/setup.go	0.00% <0.00%> (ø)
pkg/agent/run.go	0.00% <0.00%> (ø)
pkg/agent/tunnel/tunnel.go	0.00% <0.00%> (ø)
pkg/server/context.go	0.00% <0.00%> (ø)
pkg/daemons/executor/embed.go	1.13% <3.22%> (-0.56%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 998e2bb...2ad5925. Read the comment docs.