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

24th
Jun 2016
Tommy Ludwig
@shakuzen
Jun 24 2016 01:56
@dsyer I have an update on the issue I mentioned previously about files in the local Git repository for the Cloud Config Server getting deleted mysteriously. It is looking highly likely it is caused by the daily tmpwatch batch that runs on CentOS, deleting files under /tmp that haven't been accessed in 10 days.
So, not a Spring Cloud Config issue but something that might affect its users. I'm surprised no one else has ran into it and reported it before.
http://serverfault.com/questions/377348/when-does-tmp-get-cleared
Spencer Gibb
@spencergibb
Jun 24 2016 01:57
@shakuzen good call. Could set java tmp dir to somewhere out of /tmp.
Tommy Ludwig
@shakuzen
Jun 24 2016 02:03
There's also spring.cloud.config.server.git.basedir it seems
Spencer Gibb
@spencergibb
Jun 24 2016 02:03
yup
Tommy Ludwig
@shakuzen
Jun 24 2016 02:05
I would say it may be worth a note in the documentation about this potential problem (depending on your OS) but if I'm really the only one that's been affected this whole time, I'm not so sure.
Spencer Gibb
@spencergibb
Jun 24 2016 02:06
will you write an issue?
Tommy Ludwig
@shakuzen
Jun 24 2016 02:07
Sure, I'll open something up for consideration today.
pradeepkusingh
@pradeepkusingh
Jun 24 2016 04:17
@dsyer or @spencergibb , I want to reduce the hystrix matrics frequency to my rabbit mq , How can I do it ? is there any property which can reduce the hystrix matrix generation to rabbit for turbine ?
Found this property: don't know if it is for desired behavior : hystrix.command.default.metrics.rollingStats.timeInMilliseconds
Spencer Gibb
@spencergibb
Jun 24 2016 04:30
hystrix.stream.queue.sendRate defaults to 500 (which is ms)
pradeepkusingh
@pradeepkusingh
Jun 24 2016 04:30
@spencergibb : Thank you .. Thank you ...Thank you :)
Spencer Gibb
@spencergibb
Jun 24 2016 04:31
hystrix.stream.queue.gatherRate is how often metrics get put into the queue (in memory queue, before it’s shipped to rabbit). Again default is 500 ms.
pradeepkusingh
@pradeepkusingh
Jun 24 2016 04:32
should I configure both , or one of them should be good.. I don't want to send too many frequent matrics to rabbit mq as it is getting over loded..
Spencer Gibb
@spencergibb
Jun 24 2016 04:33
the queue is currently a fixed size. You’ll need to see the behavior.
pradeepkusingh
@pradeepkusingh
Jun 24 2016 04:33
ok..
BIG THANKS :) , this issue is creating problem in production and almost brought down rabbit mq couple of times
Patrick Cornelißen
@pcornelissen
Jun 24 2016 05:18
wow how many nodes do you have that generate the metrics?
pradeepkusingh
@pradeepkusingh
Jun 24 2016 05:19
total nodes are around 30..
Tommy Ludwig
@shakuzen
Jun 24 2016 06:25
@spencergibb wrote up the issue previously discussed spring-cloud/spring-cloud-config#424
Patrick Cornelißen
@pcornelissen
Jun 24 2016 09:18
Hi! Coming back to my problem of yesterday with the ssl hostname verification, I have disabled the SimpleHostRoutingFilter by setting "zuul.SimpleHostRoutingFilter.route.disable=true" which works, the ssl connection problem is gone, the mapping is visible in the /mappings endpoint. But my "MySimpleHostRoutingFilter" (see: https://github.com/orchit/zuulsslproblem/blob/master/src/main/java/com/example/MySimpleHostRoutingFilter.java) seems not to be picked up (I have added this as bean to MyConfig). The route doesn't seem to be used. No request reaches the wiremock instance. :-(
Patrick Cornelißen
@pcornelissen
Jun 24 2016 09:42
OK, I tried a few more things. As soon as I deactivate the regular SimpleHostRoutingFilter and try to get my own in place the mapping seems to die. I will now switch to http for communication between both services to get rid of this problem for now, but I think this should be addressed in spring.
litzuhsien
@litzuhsien
Jun 24 2016 10:37
@spencergibb Hi, we are using spring-cloud-netflix to develop our microservices, also using nginx as our software load balancer server. now we plan to deploy serveral edge services (zuul) instances to proxy our backend microservices. is this deployment solution in correct, if not, please give us a right direction to try...thanks
Dave Syer
@dsyer
Jun 24 2016 13:28
@litzuhsien I see nothing incorrect. What do you mean?
@pcornelissen we definitely need to support this in spring cloud. Looking at you for a pull request then?
I'm not keen to encourage ssl dodging outside tests though
litzuhsien
@litzuhsien
Jun 24 2016 13:31
Oh I see, it already runs, we will start testing our services later, thank you
Patrick Cornelißen
@pcornelissen
Jun 24 2016 16:36
@dsyer I'll have a look ;-)
Dave Syer
@dsyer
Jun 24 2016 16:37
That would be awesome
Patrick Cornelißen
@pcornelissen
Jun 24 2016 17:05
FYI I'll remove my zuulsslproblem repo as it's not a proper solution
pradeepkusingh
@pradeepkusingh
Jun 24 2016 17:26
@spencergibb : How can I filter out the hystrix stream data.. I don't want to send all info to rabbit MQ ? jsut couple of them..
Spencer Gibb
@spencergibb
Jun 24 2016 17:32
@pradeepkusingh there’s no way to do that. You’d need to disable our task and write your own.
pradeepkusingh
@pradeepkusingh
Jun 24 2016 17:35
ok , Thanks :).. I think it should be there in future.. shoud I open a issue for enhancement ?
Spencer Gibb
@spencergibb
Jun 24 2016 17:35
You can, though I’m not sure how the hystrix dashboard would react.
Dave Syer
@dsyer
Jun 24 2016 17:35
Maybe if you can explain the use case?
Spencer Gibb
@spencergibb
Jun 24 2016 17:35
hystrix stream is overwhelming their rabbitmq
Dave Syer
@dsyer
Jun 24 2016 17:36
How would you filter it, other than by sending it less frequently?
pradeepkusingh
@pradeepkusingh
Jun 24 2016 17:36
example -- If I don't want to see errro count percentage in dash board
Spencer Gibb
@spencergibb
Jun 24 2016 17:37
so not sending all of the data, not skipping whole messages
pradeepkusingh
@pradeepkusingh
Jun 24 2016 17:37
yes
Dave Syer
@dsyer
Jun 24 2016 17:37
But the dashboard will fail if it doesn't get all the fields
pradeepkusingh
@pradeepkusingh
Jun 24 2016 17:38
just want to see certain matrix . not all of them
Dave Syer
@dsyer
Jun 24 2016 17:38
You'd have t
o write your own dashboard as well then
pradeepkusingh
@pradeepkusingh
Jun 24 2016 17:38
ok..if no option.. we will explore to create dashboard
Dave Syer
@dsyer
Jun 24 2016 17:39
I seriously doubt we can support a load of optional fields in the dash
On the other hand we probably don't use all the fields in the existing dash
But it's copied directly from Netflix so maintaining patches would be expensive
Netflix dn't use it much any more
so it's tricky
pradeepkusingh
@pradeepkusingh
Jun 24 2016 17:41
ok.. then we can think an option to reomve dashboard
Dave Syer
@dsyer
Jun 24 2016 17:42
Definitely it is already optional
pradeepkusingh
@pradeepkusingh
Jun 24 2016 17:42
:)
Dave Syer
@dsyer
Jun 24 2016 17:43
So if you don't need the dashboard you can just not send any data
pradeepkusingh
@pradeepkusingh
Jun 24 2016 17:43
yes.
We are in production right now.. so I was looking for some option to diabled it dynamically
but if no option I will go with code changes
pradeepkusingh
@pradeepkusingh
Jun 24 2016 17:48
Thanks Dave and Spencer
Daco
@dacofr
Jun 24 2016 21:50
Hi SpringCloud team, i've a strange behaviour with my config server. I've connected 5 applications and all works fine. But after startup of each application i've lots of calls to my config server that i don't understand
blob
Spencer Gibb
@spencergibb
Jun 24 2016 21:51
Generally unknown calls to config server are /health calls to each app.
Daco
@dacofr
Jun 24 2016 21:52
maybe, but how to explain that all call ar done with the application name = application
Spencer Gibb
@spencergibb
Jun 24 2016 21:53
the call to /health
Daco
@dacofr
Jun 24 2016 21:53
blob
is it possible to disable this behaviour ?
Spencer Gibb
@spencergibb
Jun 24 2016 21:56
health.config.enabled=false
Daco
@dacofr
Jun 24 2016 21:57
ok, thanks Spencer
Spencer Gibb
@spencergibb
Jun 24 2016 21:58
BTW, that’s on config clients, not config server.
Daco
@dacofr
Jun 24 2016 22:00
yep, i understand. I think i will set this property in the application.properties into GIT rep and then reload all my clients
Spencer Gibb
@spencergibb
Jun 24 2016 22:01
Something will be done about it in the next release, either disabled by default or cached for a configurable amount of time (defaulting to minutes).
Daco
@dacofr
Jun 24 2016 22:02
great idea
Daco
@dacofr
Jun 24 2016 22:21
Spencer, an other question with /health for an application that is connected to eureka : actually the /health list all apps that are registered in eureka. Does it exist a solution to restrict this list only to the apps that my application use ?
Spencer Gibb
@spencergibb
Jun 24 2016 22:21
no
Daco
@dacofr
Jun 24 2016 22:24
my idea behind is to have a dynamic solution that detect all my FeignClients and extract the service name and provide a healthindicator and check that i've at least one application registered in eureka
Daco
@dacofr
Jun 24 2016 22:30
thanks for you help. Good night
Spencer Gibb
@spencergibb
Jun 24 2016 22:31
NP