OpenTelemetry module #9016

esigo · 2022-09-03T16:35:21Z

This issue is for tracking the steps needed to be taken for using OpenTelemetry.

Currently, the OpenTelemetry compilation is done in #8585, but there is still one issue for loading the module (the init container copies the needed files to a wrong path).

Here is the list of steps:

implement loading the image with the controller (fix the init container path).
integration
user-interface.
remove currently integrated bits of opentracing, jaeger, zipkin etc.

Once all done, the user will be able to send telemetry data to any backend that receives otlp-gRPC directly, or any backend through otel-collector:

graph TD
subgraph Node
   nginx
end
nginx["nginx module"]-->|otlp-gRPC| backend["backend"]

Or:

graph TD
subgraph Node
   nginx
end
nginx["nginx module"] -->|otlp-gRPC| OTEL-collector{{"Otel Collector"}} -->backend{{"backend:<br>Zipkin<br>Jaeger <br>"}}

esigo · 2022-09-03T16:35:53Z

Please assign the issue to me.

longwuyuan · 2022-09-03T16:56:06Z

You can just type & post /assign

longwuyuan · 2022-09-03T16:57:47Z

And we need to add collector code and collector compilation to the Otel bits in Makefile, I am guessing. Or is that pluggable.

longwuyuan · 2022-09-03T17:02:03Z

And remove currently integrated bits of opentracing, jaeger, zipkin etc.

esigo · 2022-09-03T17:19:01Z

/assign

esigo · 2022-09-03T17:20:23Z

And we need to add collector code and collector compilation to the Otel bits in Makefile, I am guessing. Or is that pluggable.

No, We don't need to change the Makefile. Everything is in init image.

longwuyuan · 2022-09-04T02:11:21Z

Will wait for your PR. You can begin the title with string "WIP" to indicate work-in-progress if you like. Looking forward to see your design

longwuyuan · 2022-09-05T19:12:11Z

@esigo since that PR merrged, we will get image with LD_LIB_PATH. That is good for loading. But can you write here or in that PR about the collector. Is it already compiled and ready for use with the current build ? If yes, is that also under the same /mount....... path ?

esigo · 2022-09-05T19:26:05Z

The collector has to be managed by the user. I have a sample here:
https://github.com/esigo/ingress-nginx/blob/dev2/tst/backend.sh
and here:
https://github.com/esigo/ingress-nginx/blob/dev2/tst/collector.yaml

strongjz · 2022-09-06T13:32:30Z

/triage accepted
/priority backlog
/kind feature

longwuyuan · 2022-09-08T15:23:46Z

@esigo Please test this image after path fix

docker pull gcr.io/k8s-staging-ingress-nginx/opentelemetry@sha256:aa079daa7efd93aa830e26483a49a6343354518360929494bad1d0ad3303142e
OR
docker pull gcr.io/k8s-staging-ingress-nginx/opentelemetry:v20220906-controller-v1.3.1-3-g981ce38a7

Based on your update, we can promote it or do other needful

kuzaxak · 2022-09-08T16:00:54Z

gcr.io/k8s-staging-ingress-nginx/opentelemetry:v20220906-controller-v1.3.1-3-g981ce38a7 has the same error, can't load grpc module when enabling load_module /modules_mount/etc/nginx/modules/otel/otel_ngx_module.so

esigo · 2022-09-08T16:34:50Z

Hi @longwuyuan,
Thanks, it works.
Would it be possible to show me how you promote the image? Thanks

nginx.conf:

opentelemetry image:

file on disk:

log:

kuzaxak · 2022-09-08T16:38:31Z

Strange, I have exactly the same problem.

I built my own image using this dockerfile

longwuyuan · 2022-09-08T16:41:48Z

@kuzaxak you are using image built on 6th. The new image s built on 7th.

EDIT: sorry, the comment above from me is wrong. You are using the latest. And I think esigo mentioned line 76 problem in your environ.

esigo · 2022-09-08T16:49:26Z

@kuzaxak your line 76 doesn't copy all the files that are needed.

…9242)

bixu · 2022-12-23T10:18:19Z

@esigo, wanted to express my appreciation for your work here. Very nice. I have a team at $job that would be happy to help test this in live environments if you have any pre-release artifacts to share.

esigo · 2023-01-01T12:23:23Z

pending fix on open-telemetry/opentelemetry-cpp-contrib#252 and #9472

…tes#9016) (kubernetes#9242)

esigo · 2023-03-22T19:27:00Z

/close
as completed with #9062

agronholm · 2023-03-27T07:52:15Z

Was this supposed to work on v1.7.0? Doesn't for me:

Error: exit status 1
2023/03/27 07:02:57 [emerg] 41#41: dlopen() "/modules_mount/etc/nginx/modules/otel/otel_ngx_module.so" failed (Error loading shared library /modules_mount/etc/nginx/modules/otel/otel_ngx_module.so: No such file or directory) in /tmp/nginx/nginx-cfg3874777628:10
nginx: [emerg] dlopen() "/modules_mount/etc/nginx/modules/otel/otel_ngx_module.so" failed (Error loading shared library /modules_mount/etc/nginx/modules/otel/otel_ngx_module.so: No such file or directory) in /tmp/nginx/nginx-cfg3874777628:10
nginx: configuration file /tmp/nginx/nginx-cfg3874777628 test failed

One more thing: does this require HTTPS? I don't see any option to toggle between HTTPS and HTTP.
EDIT: Looks like TLS is off by default, there is just no way to toggle that with the Helm chart settings.

fcuello-fudo · 2023-03-27T10:04:04Z

@agronholm probably you are using a daemonset instead of a deployment, I had the same issue, and found that the helm template for the daemonset is missing some settings compared to the deployment.

I have a local fix and I'm preparing a pull request but I need to sign CLA, etc.

agronholm · 2023-03-27T10:04:48Z

No, I'm using deployment.

fcuello-fudo · 2023-03-27T10:06:23Z

@agronholm , did you add:

controller:
  opentelemetry:
    enabled: true

in addition to:

controller:
  config:
    enable-opentelemetry: "true"

?

agronholm · 2023-03-27T10:10:46Z

I didn't. I tried that and my controller starts up now. Is this documented somewhere?

fcuello-fudo · 2023-03-27T10:11:36Z

@agronholm , no, I also found that by debugging the chart. There are also many mistakes in the documentation :(

agronholm · 2023-03-27T10:12:36Z

Okay. Thanks for the info! What mistakes did you find in the documentation?

hargut · 2023-03-30T15:00:23Z

@fcuello-fudo
Thank you for the hint & fix for the DaemonSet issue, I was able to successfully use the patched version from your repository. 👍

fcuello-fudo · 2023-03-30T15:03:08Z

Ref.: #9792

eshizhan · 2023-04-15T13:16:28Z

How can I enable opentelemetry without helm, only using kubectl apply command with some yaml manifests. I can not find any way to do this, is support now?

premjha9625 · 2023-04-17T08:43:51Z

Hi , i want to export ingress controller traces to azure app insights. how can i achieve that using the open telemetry provided by ingress controller. any solution ??

sereinity · 2023-04-27T14:19:18Z

Hello,

I wanted to give it a try, and I got a huge CPU usage issue, every core are saturated by nginx. I simplified my setup to reduce external causes, I deleted ALL my ingress objects, and deploy the helm chart with this values.yaml:

controller:
  config:
    enable-opentelemetry: "true"
    zipkin-collector-host: otel-collector.opentelemetry.svc
  opentelemetry:
    enabled: true
  replicaCount: 2
  service:
    loadBalancerIP: 1.1.1.1 # a public IP

Yes, otel is

And there is what a get from a top in one of the pod (the other looks the same):

Mem: 6941004K used, 25938868K free, 55596K shrd, 275676K buff, 4664064K cached
CPU0:  83% usr  16% sys   0% nic   0% idle   0% io   0% irq   0% sirq
CPU1: 100% usr   0% sys   0% nic   0% idle   0% io   0% irq   0% sirq
CPU2: 100% usr   0% sys   0% nic   0% idle   0% io   0% irq   0% sirq
CPU3: 100% usr   0% sys   0% nic   0% idle   0% io   0% irq   0% sirq
CPU4: 100% usr   0% sys   0% nic   0% idle   0% io   0% irq   0% sirq
CPU5: 100% usr   0% sys   0% nic   0% idle   0% io   0% irq   0% sirq
CPU6: 100% usr   0% sys   0% nic   0% idle   0% io   0% irq   0% sirq
CPU7: 100% usr   0% sys   0% nic   0% idle   0% io   0% irq   0% sirq
Load average: 10.13 10.66 11.38 14/1028 357
  PID  PPID USER     STAT   VSZ %VSZ CPU %CPU COMMAND
   75    26 www-data S     154m   0%   1  13% nginx: cache manager process
   30    26 www-data S     168m   0%   0  13% nginx: worker process
   39    26 www-data S     168m   0%   5  13% nginx: worker process
   35    26 www-data S     168m   0%   7  13% nginx: worker process
   32    26 www-data S     168m   0%   3  12% nginx: worker process
   33    26 www-data S     168m   0%   1  11% nginx: worker process
   36    26 www-data S     168m   0%   3   9% nginx: worker process
   34    26 www-data S     168m   0%   1   8% nginx: worker process
   31    26 www-data S     168m   0%   4   8% nginx: worker process
    7     1 www-data S     737m   2%   1   0% /nginx-ingress-controller --publish-service=ingress/ingress-nginx-controller --election-id=ingress-nginx-leader --controller-class=k8s.io/ingress-nginx --ingress-class=nginx --configmap=ingress/ingress-nginx-controller --validat
  346     0 www-data S     257m   1%   3   0% top
  352     0 www-data R     257m   1%   1   0% top
   26     7 www-data S     156m   0%   1   0% nginx: master process /usr/bin/nginx -c /etc/nginx/nginx.conf
    1     0 www-data S      216   0%   7   0% /usr/bin/dumb-init -- /nginx-ingress-controller --publish-service=ingress/ingress-nginx-controller --election-id=ingress-nginx-leader --controller-class=k8s.io/ingress-nginx --ingress-class=nginx --configmap=ingress/ingress-ngin

If I comment enable-opentelemetry: "true" I get a normal CPU usage:

Mem: 7518752K used, 25361120K free, 94200K shrd, 285928K buff, 5047600K cached
CPU0:  40% usr  20% sys   0% nic  40% idle   0% io   0% irq   0% sirq
CPU1:   0% usr  16% sys   0% nic  83% idle   0% io   0% irq   0% sirq
CPU2:  20% usr   0% sys   0% nic  80% idle   0% io   0% irq   0% sirq
CPU3:  20% usr   0% sys   0% nic  80% idle   0% io   0% irq   0% sirq
CPU4:   0% usr  20% sys   0% nic  80% idle   0% io   0% irq   0% sirq
CPU5:   0% usr   0% sys   0% nic 100% idle   0% io   0% irq   0% sirq
CPU6:   0% usr   0% sys   0% nic 100% idle   0% io   0% irq   0% sirq
CPU7:   0% usr   0% sys   0% nic 100% idle   0% io   0% irq   0% sirq
Load average: 5.22 9.04 10.70 5/1109 633
  PID  PPID USER     STAT   VSZ %VSZ CPU %CPU COMMAND
  367    26 www-data S     167m   0%   4   3% nginx: worker process
    7     1 www-data S     737m   2%   1   0% /nginx-ingress-controller --publish-service=ingress/ingress-nginx-controller --election-id=ingress-nginx-leader --controller-class=k8s.io/ingress-nginx --ingress-class=nginx --configmap=ingress/ingress-nginx-controller --validat
  346     0 www-data S     257m   1%   4   0% top
  628     0 www-data R     257m   1%   2   0% top
  368    26 www-data S     167m   0%   3   0% nginx: worker process
  363    26 www-data S     167m   0%   4   0% nginx: worker process
  365    26 www-data S     167m   0%   0   0% nginx: worker process
  369    26 www-data S     167m   0%   1   0% nginx: worker process
  362    26 www-data S     167m   0%   1   0% nginx: worker process
  366    26 www-data S     167m   0%   1   0% nginx: worker process
  364    26 www-data S     167m   0%   4   0% nginx: worker process
   26     7 www-data S     156m   0%   2   0% nginx: master process /usr/bin/nginx -c /etc/nginx/nginx.conf
  370    26 www-data S     154m   0%   3   0% nginx: cache manager process
    1     0 www-data S      216   0%   7   0% /usr/bin/dumb-init -- /nginx-ingress-controller --publish-service=ingress/ingress-nginx-controller --election-id=ingress-nginx-leader --controller-class=k8s.io/ingress-nginx --ingress-class=nginx --configmap=ingress/ingress-ngin

Despise the huge CPU usage, I get some traces (the stack is ingress + otel + a collector + tempo + grafana).

I may have miss something, does someone have an idea?

svmaris · 2023-04-28T06:38:16Z

@sereinity I had the same issue. This was fixed when I set otel-schedule-delay-millis to something like 5000 and otel-max-export-batch-size to 512 in the ConfigMap. The defaults are causing every span to be sent to the collector without any batching/delay, which results in huge CPU usage.

strongjz · 2023-06-08T15:23:56Z

Otel module has been completed, I am going to close this issue.

If folks have issues with the modules, please open separate issues tickets for those. thank you.

/close

strongjz · 2023-06-08T15:24:29Z

@sereinity @svmaris i have created this ticket #10063

deejgregor · 2023-06-09T01:48:14Z

@sereinity @svmaris as you might have seen in #10063, the performance issues were fixed with defaults changes in #9978 (see ac9a507). This was released in v1.8.0 of the controller / v4.7.0 of the Helm chart.

deejgregor · 2023-06-09T01:55:31Z

Hi , i want to export ingress controller traces to azure app insights. how can i achieve that using the open telemetry provided by ingress controller. any solution ??

@premjha9625 I haven't tried this, but it looks like the way to do this today is to send your ingress-nginx trace to OpenTelemetry Collector and use the Azure Monitor Exporter as mentioned here: https://learn.microsoft.com/en-us/azure/azure-monitor/faq#can-i-use-the-opentelemetry-collector-

Angelin01 · 2023-06-21T21:10:27Z

Wonderful work! However, the documentation could be improved a bit still. For example.

# specifies the name to use for the server span
opentelemetry-operation-name

We are also given this example elsewhere: opentelemetry-operation-name: "HTTP $request_method $service_name $uri"

But what are the available variables for this? Is this information available? If so, where can I find it?

Either way, I could create another issue and even take a crack at this if someone points me in the right direction. My random code dives haven't been fruitful, unfortunately.

longwuyuan · 2023-06-22T00:42:55Z

@Angelin01 maybe you can reach to @esigo on the ingress-nginx-dev channel of kubernetes.slack.com. Register at slack.k8s.io if required.

lsunsi · 2023-11-08T14:14:47Z

@eshizhan Did you find out how to enable opentelemetry using the bare apply method instead of helm method? I have the controller deployed and I can't find anyway to pass this "controller: true" people are talking about.

k8s-ci-robot added needs-triage Indicates an issue or PR lacks a `triage/foo` label and requires one. needs-kind Indicates a PR lacks a `kind/foo` label and requires one. needs-priority labels Sep 3, 2022

k8s-ci-robot assigned esigo Sep 3, 2022

esigo mentioned this issue Sep 4, 2022

fix LD_LIBRARY_PATH for opentelemetry #9017

Merged

11 tasks

longwuyuan mentioned this issue Sep 5, 2022

Working OpenTelemetry sidecar (without base nginx image) #8735

Closed

9 tasks

esigo mentioned this issue Sep 6, 2022

fix otel init_module #9028

Merged

11 tasks

longwuyuan mentioned this issue Sep 8, 2022

An error occurred between opentelemetry modules #8437

Closed

k8s-ci-robot pushed a commit that referenced this issue Nov 10, 2022

add containerSecurityContext to extraModules init containers (#9016) (#…

726d7e6

…9242)

This was referenced Nov 10, 2022

OpenTelemetry static linking #9286

Merged

update OpenTelemetry image #9308

Merged

jaehnri pushed a commit to jaehnri/ingress-nginx that referenced this issue Jan 2, 2023

add containerSecurityContext to extraModules init containers (kuberne…

2b80a1c

…tes#9016) (kubernetes#9242)

esigo mentioned this issue May 29, 2023

Add legacy to OpenTelemetry migration doc #10011

Merged

10 tasks

strongjz closed this as completed Jun 8, 2023

strongjz mentioned this issue Jun 8, 2023

update defaults for the otel collector to reduce CPU/Memory #10063

Closed

OpenTelemetry module #9016

OpenTelemetry module #9016

Comments

esigo commented Sep 3, 2022 • edited Loading

esigo commented Sep 3, 2022

longwuyuan commented Sep 3, 2022

longwuyuan commented Sep 3, 2022

longwuyuan commented Sep 3, 2022

esigo commented Sep 3, 2022

esigo commented Sep 3, 2022

longwuyuan commented Sep 4, 2022

longwuyuan commented Sep 5, 2022 • edited Loading

esigo commented Sep 5, 2022

strongjz commented Sep 6, 2022

longwuyuan commented Sep 8, 2022 • edited Loading

kuzaxak commented Sep 8, 2022

esigo commented Sep 8, 2022

kuzaxak commented Sep 8, 2022 • edited Loading

longwuyuan commented Sep 8, 2022 • edited Loading

esigo commented Sep 8, 2022

bixu commented Dec 23, 2022

esigo commented Jan 1, 2023

esigo commented Mar 22, 2023 • edited Loading

agronholm commented Mar 27, 2023 • edited Loading

fcuello-fudo commented Mar 27, 2023

agronholm commented Mar 27, 2023

fcuello-fudo commented Mar 27, 2023

agronholm commented Mar 27, 2023

fcuello-fudo commented Mar 27, 2023 • edited Loading

agronholm commented Mar 27, 2023

hargut commented Mar 30, 2023

fcuello-fudo commented Mar 30, 2023

eshizhan commented Apr 15, 2023

premjha9625 commented Apr 17, 2023

sereinity commented Apr 27, 2023

svmaris commented Apr 28, 2023

strongjz commented Jun 8, 2023

strongjz commented Jun 8, 2023

deejgregor commented Jun 9, 2023 • edited Loading

deejgregor commented Jun 9, 2023 • edited Loading

Angelin01 commented Jun 21, 2023

longwuyuan commented Jun 22, 2023

lsunsi commented Nov 8, 2023 • edited Loading

esigo commented Sep 3, 2022 •

edited

Loading

longwuyuan commented Sep 5, 2022 •

edited

Loading

longwuyuan commented Sep 8, 2022 •

edited

Loading

kuzaxak commented Sep 8, 2022 •

edited

Loading

longwuyuan commented Sep 8, 2022 •

edited

Loading

esigo commented Mar 22, 2023 •

edited

Loading

agronholm commented Mar 27, 2023 •

edited

Loading

fcuello-fudo commented Mar 27, 2023 •

edited

Loading

deejgregor commented Jun 9, 2023 •

edited

Loading

deejgregor commented Jun 9, 2023 •

edited

Loading

lsunsi commented Nov 8, 2023 •

edited

Loading