Evaluation of Istio 1.16

strekm commented 1 year ago

Status Waiting for @kyma-project/observability to check dashboard integration

Description

Prepare gardener cluster with Istio 1.16 installation and execute manual tests. Prepare gardener cluster with latest kyma main and upgrade istio to 1.16. Execute manual tests. Share cluster with other teams to evaluate integration with other kyma components. Check proxy reset job. Check grafana dashboard changes. Production profile should be use in both cases.

ACs:

[x] Istio 1.16 installs
[x] Kyma can be upgraded to Istio 1.16
[x] istio 1.16 installs with cni and w/o cni enabled
[x] manual tests executed for both gardener clusters
[x] execute manual perf tests

On Kyma eval profile

Allow

running (1m00.2s), 000/100 VUs, 38951 complete and 0 interrupted iterations
default ✓ [======================================] 100 VUs  1m0s

 ✗ SuccessFull calls
  ↳  99% — ✓ 38946 / ✗ 5

 checks.....................: 99.98% ✓ 38946      ✗ 5    
 data_received..............: 4.8 MB 79 kB/s
 data_sent..................: 1.5 MB 24 kB/s
 http_req_blocked...........: avg=517.54µs min=191ns   med=431ns    max=650.53ms p(90)=561ns    p(95)=635ns   
 http_req_connecting........: avg=15.32µs  min=0s      med=0s       max=35.15ms  p(90)=0s       p(95)=0s      
 http_req_duration..........: avg=153.35ms min=1.67ms  med=122.82ms max=879.94ms p(90)=284.63ms p(95)=311.06ms
 http_req_failed............: 0.01%  ✓ 5          ✗ 38946
 http_req_receiving.........: avg=3.01ms   min=16.86µs med=750.12µs max=65.98ms  p(90)=8.61ms   p(95)=13.37ms 
 http_req_sending...........: avg=468.98µs min=26.12µs med=74.47µs  max=123.17ms p(90)=148.8µs  p(95)=849.2µs 
 http_req_tls_handshaking...: avg=410.66µs min=0s      med=0s       max=591.29ms p(90)=0s       p(95)=0s      
 http_req_waiting...........: avg=149.87ms min=0s      med=119.15ms max=879.75ms p(90)=280.04ms p(95)=307.06ms
 http_reqs..................: 38951  647.287058/s
 iteration_duration.........: avg=154.21ms min=1.86ms  med=123.35ms max=1.09s    p(90)=285.04ms p(95)=311.66ms
 iterations.................: 38951  647.287058/s
 vus........................: 100    min=100      max=100
 vus_max....................: 100    min=100      max=100

Oauth2

running (1m01.5s), 000/100 VUs, 3251 complete and 0 interrupted iterations
default ✓ [======================================] 100 VUs  1m0s

     ✓ SuccessFull calls

     █ setup

     checks.....................: 100.00% ✓ 3251      ✗ 0    
     data_received..............: 908 kB  15 kB/s
     data_sent..................: 877 kB  14 kB/s
     http_req_blocked...........: avg=7.31ms   min=254ns    med=536ns    max=553.54ms p(90)=801ns    p(95)=1.07µs
     http_req_connecting........: avg=380.77µs min=0s       med=0s       max=50.2ms   p(90)=0s       p(95)=0s    
     http_req_duration..........: avg=1.85s    min=118.3ms  med=1.8s     max=5.1s     p(90)=2.77s    p(95)=3.19s 
     http_req_failed............: 0.00%   ✓ 0         ✗ 3252 
     http_req_receiving.........: avg=395.31µs min=22.59µs  med=73.43µs  max=19.72ms  p(90)=723.59µs p(95)=2.17ms
     http_req_sending...........: avg=717.73µs min=57.7µs   med=171.02µs max=32.7ms   p(90)=1.6ms    p(95)=2.85ms
     http_req_tls_handshaking...: avg=5.84ms   min=0s       med=0s       max=484.89ms p(90)=0s       p(95)=0s    
     http_req_waiting...........: avg=1.85s    min=117.93ms med=1.79s    max=5.1s     p(90)=2.77s    p(95)=3.19s 
     http_reqs..................: 3252    52.873741/s
     iteration_duration.........: avg=1.86s    min=169.03ms med=1.8s     max=5.1s     p(90)=2.79s    p(95)=3.19s 
     iterations.................: 3251    52.857483/s
     vus........................: 35      min=35      max=100
     vus_max....................: 100     min=100     max=100

[x] security scans
[x] upgrading istio to 1.16 does not spoil integration with other components (kiali, tracing, prometheus) (huskies)
[x] validate if workload is added to service mesh during reconciliation
[x] test GKE cni plugin | @barchw --> works with no overrides, installed in kube-system

Reasons

Keeping Istio up to date

Attachments https://istio.io/latest/news/releases/1.16.x/announcing-1.16/ https://github.com/kyma-project/kyma/issues/16231

videlov commented 1 year ago

test integration, slack topic

videlov commented 1 year ago

ToDos/Leftovers:

execute load/perf tests
check if huskies / other teams are done with the 1.16 test cluster (keep it up-and-running): https://dashboard.garden.canary.k8s.ondemand.com/namespace/garden-goatz/shoots/istio-1-16/
not sure if it worked for GKE with cni, check https://console.cloud.google.com/kubernetes/clusters/details/europe-central2-a/istio-1-16/details?authuser=0&project=sap-se-cx-kyma-goat

rakesh-garimella commented 1 year ago

Tested following by manually installing the istio 1.16 on the cluster:

Stress test with prometheus has no surprises
Kiali coming out of the box gives warning that it might not be compatible. But checking the graphs looks fine.
Installing Kiali using examples also works fine and the warning is not present.
Grafana dashboards for the istio are fine
Otel example works fine

kyma-project / kyma

Evaluation of Istio 1.16 #16157