Jaeger: Setup Jaeger for N services

[Disturbing]

Is it possible to setup Jaeger to only show a specific whitelist of services?

I tried googling it around and couldn't find it.

The idea is to:

Setup Jaeger A for service A,B,C and D only
Setup Jaeger B for service Z,Y,W only

It would be best to be able to do this with kubernetes labels as well :).

Thanks in advance!

[yurishkuro]

This is a question about environments isolation, I don't think it requires any specific support from Jaeger.

[Disturbing]

Maybe I dont understand how jaeger works? Is it possible to setup two
jaeger instances in two different namespaces and it confines itself to
those services on Kubernetes? Maybe I should be asking this question
somewhere else?
On Sat, Apr 28, 2018 at 9:52 PM Yuri Shkuro [e-mail]
wrote:

This is a question about environments isolation, I don't think it requires
any specific support from Jaeger.

—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
https://github.com/jaegertracing/jaeger/issues/795#issuecomment-385177484,
or mute the thread
https://github.com/notifications/unsubscribe-auth/ACQev5AOuDXND3epzbKi-Dyr0C-p8IVvks5ttHQWgaJpZM4TrSML
.

[yurishkuro]

Jaeger does not do any automatic discovery of services. When you configure Jaeger clients in your applications you tell them where to send captured traces. If you are running jaeger-agent as a sidecar or telling clients to submit directly to jaeger-collector, then you're in full control of which Jaeger backend the data is going to end up.

It's fine to ask questions here, btw. The other options are mailing list and chat https://gitter.im/jaegertracing/Lobby.

[jpkrohling]

Complementing what @yurishkuro said, this can (and should, IMO) be done at the infra layer. In Kubernetes, it would mean isolating your services into "namespaces". In OpenShift, it would be in "projects".

Note that you cannot use DaemonSet on such a setup, as they run at the node level without caring about which namespaces the running pods belong to: you should either use the agent as sidecar or make the client talk directly to the collector.

[Disturbing]

Okay -- Thanks for all the quick replies guys --

Right now I'm using istio and jaeger's all-in-one template. Jaeger is
deployed in logging and my services in default and jaeger is a sidecar to
them all if I'm not mistaken.

I'm looking potentially deploy jaeger to dev and qa namespaces as a sidecar
and when I view Jaeger is either qa or dev, you would only see the services
within that namespace.

--

Better yet, If I can append some sort of header where Jaeger only listens
and records to services with that header, that would be an amazing win for
my requirements.

Essentially, I'm trying to deploy jaeger as a service where it's only
listening to the pool of services that the client cares about, not the
entire ecosystem which could be other client's applications.

On Mon, Apr 30, 2018 at 1:48 PM Juraci Paixão Kröhling <
[email protected]> wrote:

Complementing what @yurishkuro https://github.com/yurishkuro said, this
can (and should, IMO) be done at the infra layer. In Kubernetes, it would
mean isolating your services into "namespaces"
https://kubernetes.io/docs/concepts/overview/working-with-objects/namespaces/.
In OpenShift, it would be in "projects"
https://docs.openshift.org/latest/dev_guide/projects.html.

Note that you cannot use DaemonSet on such a setup, as they run at the
node level without caring about which namespaces the running pods belong
to: you should either use the agent as sidecar or make the client talk
directly to the collector.

—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
https://github.com/jaegertracing/jaeger/issues/795#issuecomment-385317568,
or mute the thread
https://github.com/notifications/unsubscribe-auth/ACQev2kbbbLgmRpBt3pJh424T3Ovm1hYks5ttqWxgaJpZM4TrSML
.