These are chat archives for spring-cloud/spring-cloud

1st
Dec 2015
Kirill Tolkachev
@lavcraft
Dec 01 2015 10:50
Hello. Can i reg my application in Eureka with custom port?
Dave Syer
@dsyer
Dec 01 2015 10:50
You tried it and it doesn't work?
What do you mean?
Kirill Tolkachev
@lavcraft
Dec 01 2015 10:55
No, thats work with default config very well, but i use Docker, and have next problem: docker use port mapping 0.0.0.0:RANDOM_PORT->8080, application reg in eureka with 8080 port. I would specify PROPERTY_X=RANDOM_PORT_FROM_EXTERNAL_SYSTEM for say "It is my real port! Use it!"
In docker container i use env property $PORT0 ( mesos with framework marathon add to container $PORT[0-N] property)
Like this : java -Xmx{{manifest.acquiring.xmx}}M -DEUREKA_SERV={% for host in groups['mesos-slave'] %}http://{{host}}:8761/eureka/,{% endfor %} -jar $MESOS_SANDBOX/app-{{manifest.acuiring.version}}.jar --server.port=$PORT0
Dave Syer
@dsyer
Dec 01 2015 11:01
I'm not sure what that all means. But you set server.port so it should get sent to eureka correctly by default.
Is it the eureka server port that's the issue (8761 is hard-coded in that command line)?
Kirill Tolkachev
@lavcraft
Dec 01 2015 11:12
@dsyer thank you. eureka.intance.nonSecurePort works for me.
Kirill Tolkachev
@lavcraft
Dec 01 2015 11:17
use java -jar ... --eureka.instance.nonSecurePort=$PORT0 In marathon manifest with internal static port. It works fine
Dave Syer
@dsyer
Dec 01 2015 11:18
Should work either way. At least if you use the latest Spring Cloud bits.
Kirill Tolkachev
@lavcraft
Dec 01 2015 12:22
We have been using spring from release 1.3.0 :+1: thats why we donhave any problems. Mm... There is only a problem with hostname/ip resolving in docker container, but this is another story ...:)
Kirill Tolkachev
@lavcraft
Dec 01 2015 12:30
we using eureka.instance.hostname=<hostname> property for resolve this problem
Joao Duraes
@joaoduraes
Dec 01 2015 16:37
Hello, @dsyer… yesterday I mentioned an issue I was having with feign clients… I have created two sample projects in my github that replicate my problem…
the 2 projects are - https://github.com/joaoduraes/feigntests-projectwithclient and https://github.com/joaoduraes/feigntests-projectthatrequiresclient
when trying to boot up the project that requires the client I get a No qualifying bean of type [com.joaoduraes.projectwithclient.SomeClient]
Dave Syer
@dsyer
Dec 01 2015 16:41
That's because your @EnableFeignClients doesn't scan the packages in the other project
Joao Duraes
@joaoduraes
Dec 01 2015 16:42
hmm but shouldn’t it?
Dave Syer
@dsyer
Dec 01 2015 16:42
It only scans packages below where it finds itself by default
Just like a regular @ComponentScan
Joao Duraes
@joaoduraes
Dec 01 2015 16:43
shouldn’t it look to the packages in the @ComponentScan?
Dave Syer
@dsyer
Dec 01 2015 16:43
If you want to scan other packages you have to ask it to
Joao Duraes
@joaoduraes
Dec 01 2015 16:43
but how?
Dave Syer
@dsyer
Dec 01 2015 16:43
No, it doesn't know anything about a @ComponentScan
Joao Duraes
@joaoduraes
Dec 01 2015 16:43
oh, I see
I can define the basePackages for the @EnableFeignClients also
Dave Syer
@dsyer
Dec 01 2015 16:43
Yes
It's consistent with other scanning annotations
I think
Joao Duraes
@joaoduraes
Dec 01 2015 16:44
couldn’t it use the packages declared in the @ComponentScan by default?
I think it would make sense
Dave Syer
@dsyer
Dec 01 2015 16:46
Don't know
If you can find any other examples of scanning annotations that do that we can copy them
AFAIK they don't
Joao Duraes
@joaoduraes
Dec 01 2015 16:48
what you are saying is that the @EnableFeignClients isn’t aware of what’s on the @ComponentScan, right?
Dave Syer
@dsyer
Dec 01 2015 16:49
Yes
Joao Duraes
@joaoduraes
Dec 01 2015 16:51
yes, it’s probably not an easy thing to implement
anyway, the issue I had is solved now
thanks for the input :)
checketts
@checketts
Dec 01 2015 17:46
@spencergibb Have you had a chance to check out spring-cloud/spring-cloud-config#272 ? The Encrypt endpoint is is broken for us in the meantime (since it doesn't properly encode the prefix {key:someKey}mysecret )
The PR and test are complete. Please let me know if you need any additional stuff from me or @westse (like signing some contributor agreement)
Dave Syer
@dsyer
Dec 01 2015 17:47
Did you do that?
checketts
@checketts
Dec 01 2015 17:48
Do what? Sign the contributor agreement?
Dave Syer
@dsyer
Dec 01 2015 17:49
Yes
checketts
@checketts
Dec 01 2015 17:52
Yes I signed it and so did @westse
Dave Syer
@dsyer
Dec 01 2015 17:52
It's a bug in 1.0.x right?
But you sent the PR to master
(Thanks for the CLA)
checketts
@checketts
Dec 01 2015 17:53
We are running Brixton milestone 3, so I hadn't tried it against the 1.0.x branch. I suspect it was broken in that branch too.
I'm fine if it is merely fixed in master.
Dave Syer
@dsyer
Dec 01 2015 17:54
I'll see if I can port the change
checketts
@checketts
Dec 01 2015 17:54
Sweet, thanks
Dave Syer
@dsyer
Dec 01 2015 17:54
But then you won't get the author attribution
checketts
@checketts
Dec 01 2015 17:54
That will be fine
Dave Syer
@dsyer
Dec 01 2015 18:15
That change is in master now
(and 1.0.x)
checketts
@checketts
Dec 01 2015 18:18
Thanks
ccit-spence
@ccit-spence
Dec 01 2015 21:05
I am noticing that when you hit the /health endpoint it is causing a call to fetch the config client. What is that call for?
Dave Syer
@dsyer
Dec 01 2015 21:51
It's the config server health check
You can disable it if you don't like it
But if it's DOWN then you can't refresh the config
ccit-spence
@ccit-spence
Dec 01 2015 22:00
ok, thought it was something like that. I was noticing it taking around 2s compared to the version call at 5ms to 12ms
Dave Syer
@dsyer
Dec 01 2015 22:05
You probably have an issue in the config server then
Slow disk? Low memory?
ccit-spence
@ccit-spence
Dec 01 2015 22:06
Yeah, will do some research. Could be network related as well.
Has to be something since it is such a glaring difference
I will give credit to the Hystrix Dashboard. Things like that on a per call basis are much easier to spot