Zipkin: Can zipkin-sever be distributed deployment!

[luoyongjiee]

Hello
If I have a lot of spans to consume(kafka+cassandra), one zipkin-sever will take much time to insert into cassandra! Too slowly! Can zipkin-sever be distributed deployment!
Any good ideas! Thank you!

[adriancole]

yes, this is related to your other question, where the answer implied you can run multiple instances of zipkin #1159

When we released the new server, we clarified this topic on the mailing list, too, which you are free to join https://groups.google.com/forum/#!topic/zipkin-user/VtyR3GD-0Fo

[dragontree101]

if i want to run multiple instances of zipkin, i only need to start zipkin server, and these zipkin server has the same config?

@adriancole if i run multiple instances of zipkin on multiple ip, does these ip both have web ui? can i start up multiple instances of zipkin-collector and only one zipkin-query and one zipkin-web-ui?

[adriancole]

@dragontree101 there's not much point in creating a separate JVM for the UI, as the UI is just javascript and configuration. Some host the UI on nginx, but we don't support this at the moment.

Configuration aside, there's no state shared between the api/ui components and the collector ones.

This means yes, you can choose to send api/ui traffic to different servers than you send spans to.

[dragontree101]

i don't understand your mean, sorry.

i start zipkin-server by command java -jar zipkin-server-1.0.0-exec.jar and i can visit web in localhost:9411, if i start up 3 zipkin-server in different server, i think i can visit web in zipkin-server-1:9411, zipkin-server-2:9411, zipkin-server-3:9411

can i only use one zipkin-server-1:9411 and close others?

[adriancole]

yep. they don't coordinate or share state between each other. (ps use a
later version like 1.2.1 :) )

think of them like you would a normal stateless api server.