ScaleCube/Lobby

https://github.com/scalecube/scalecube-services ⚛ High-Speed ⚛ Cloud-Native ⚛ Reactive Microservices ⚓

segabriel

@segabriel

Hi we use rsocket and we should use the same versions of netty and reactor-netty.
So compare our versions https://github.com/scalecube/scalecube-services/blob/develop/pom.xml

Eugene Utkin

@eutkin

ok, i will check my dependencies, thank you for answer

Eugene Utkin

@eutkin

I found a mistake. The newer version is due to spring-boot-starter-reactor-netty. Thank you all

Swapnil Pandey

@swapnil-pandey

After I add encoder/decoders I still am getting the same error.

static class MetaDecoder implements MetadataDecoder {
        @Override
        public Object decode(ByteBuffer buffer) {
            return buffer;
        }
    }

    static class MetaEncoder implements MetadataEncoder {
        @Override
        public ByteBuffer encode(Object metadata) {
            ByteBuffer buffer = ByteBuffer.allocate(0);
            return buffer;
        }
    }

2019-09-12 08:36:52.572 WARN 24 --- [-cluster-3000-1] i.s.cluster.metadata.MetadataStoreImpl : Failed getting GetMetadataResp[17f7ae862c524ec6-1568277258724] from ip1:3000 within 3000 ms [at 17f7ae862c524ec6@ip2:3000], cause : java.util.concurrent.TimeoutException: Did not observe any item or terminal signal within 3000ms in 'source(MonoDefer)' (and no fallback has been configured)

Swapnil Pandey

@swapnil-pandey

@artem-v @ronenhamias

Artem Vysochyn

@artem-v

@swapnil-pandey I will take a look today

Artem Vysochyn

@artem-v

Share gist somewhere of program u're running and having troubles with

Also take a look -- https://gist.github.com/artem-v/32b3af36361f3c518109fdacaa74820e, in this example I have applied those metadata codecs which u had posted in the gitter and .. it works.

Swapnil Pandey

@swapnil-pandey

The error which I was talking about does not come immediately but rather after a wait of ~20 seconds

Swapnil Pandey

@swapnil-pandey

@artem-v sent you the gist in private chat

Swapnil Pandey

@swapnil-pandey

any luck figuring out the cause of the error?

Artem Vysochyn

@artem-v

I didn't look at it yet
Was busy on main project
Let me see

Artem Vysochyn

@artem-v

@ronenhamias how you see if I extract scalecube-services-discovery into separate repository? I have noticed that we bumping versions of sc-services but cause for this not api updates of services themselves, but changes in the discovery module when upgrade scalecube-cluster.

Ronen

@ronenhamias

@artem-v for now scalecube services discovery is really tied to scalecube services

unless we generalize this component for general use i dont see value of seperation

Artem Vysochyn

@artem-v

unless we generalize this component for general use

ok
let's do that and then separate

Swapnil Pandey

@swapnil-pandey

https://github.com/swapnil-pandey/scaecubetest
Just clone this repo and change the seed node and host ip accordingly.

when i run this on 2 nodes (ie cluster size =2), then if the network at anytime gets choppy, i get this warning:

2019-10-01 15:40:15.673 WARN 9838 --- [-cluster-3000-1] i.s.cluster.metadata.MetadataStore : Failed getting GetMetadataResp[132aef8ead54697-1569924532000] from 192.168.1.177:3000 within 3000 ms [at 132aef8ead54697@192.168.1.78:3000], cause : java.util.concurrent.TimeoutException: Did not observe any item or terminal signal within 3000ms in 'source(MonoDefer)' (and no fallback has been configured)
just want to know if it effects the stability of the code or i can hide this warning through any way.

Ronen

@ronenhamias

this warning just say metadata update failed

since you are not using metadata it has no effect

Swapnil Pandey

@swapnil-pandey

ok. after some time all i see in the logs are these errors, so i was worried that it might affect the performance in some way.

Ronen

@ronenhamias

updating metadata is a diffrent path

when gossip message arrives with some new node

it try to update metadata

if there is temp network issue it might fail

but its ok since eventually if there is network it should be updated

gossip is spread across nodes

but update metadata is a request response

so its a diffrent path

Swapnil Pandey

@swapnil-pandey

One thing I have noticed is that I have a cluster of 7 nodes. All on aws ec2 instances and connected to the same seed node(which is one of the ec2 instance). I am logging the cluster size cluster members on each node every 3-4 seconds.
Sometimes it happens that the cluster size on the node goes from 7 to 1(ie, itself) abruptly. It takes some time to scale back to 7.

What may be the cause of that? all nodes are ec2 instances so I dont think network can be a problem:!

Swapnil Pandey

@swapnil-pandey

Has anyone integrated prometheus with scalecube?

Ronen

@ronenhamias

did you verify the network? since it sounds like you suffer from some network configuration
we are running on ec2 instances all the time in prod with no issues

Swapnil Pandey

@swapnil-pandey

I am running the set-up on ec2 instances which makes me think there should not be a network problem :\

Ronen

@ronenhamias

hard to say anything about your environment without detailed logs
you asked if there is known issue so the answer is no.

Rajesh Kumar K V

@catchme1412

I tried removing memberHost config from one of the cluster. But it still fails to communicate. Can you please try running the proxy and agent in two different nodes and see whether the code still works?

Artem Vysochyn

@artem-v

@catchme1412
Short reminder, memberHost is how other nodes would see u and try to connect to u

By removing it u falls back to /etc/hostname (and consequently) to /etc/hosts

Rajesh Kumar K V

@catchme1412

I understood what you are telling, but sorry to say I am still not sure how make my cluster code to communicate each other when running on different IPs

Eugene Utkin

@eutkin

Hey, everybody. Can you tell us how to set up ci for scalecube-spring-boot-starter? Do you have any instructions? Or by analogy with other projects?

Artem Vysochyn

@artem-v

Hello
I suppose by analogy with other projects

Eugene Utkin

@eutkin

ok, thx

Ronen

@ronenhamias

@eutkin please take a look on the project https://github.com/scalecube/scalecube-travis-ci

Eugene Utkin

@eutkin

ok, thx

Swapnil Pandey

@swapnil-pandey

Can a cluster have multiple seeds?

Ronen

@ronenhamias

@swapnil-pandey yes it can

seed can be any node in the cluster

seed is also relative to the node that joins the cluster

so basically seed is 1 or more nodes that their address is known when node joins the cluster

RussellWilby

@RussellWilby

Hi, the latest version of scalecube-services-gateway-rsocket (2.6.0) seems to not work with the latest scalecube-services (2.8.9) release in mvnrepository. I get ClassNotFoundException: io.rsocket.frame.decoder.PayloadDecoder. RSocket version issue ?

Artem Vysochyn

@artem-v

@RussellWilby take latest gateway 2.8.13. Artifact is here -- https://repo1.maven.org/maven2/io/scalecube/scalecube-services-gateway-netty/2.8.13/.

Where communities thrive

People

Repo info

Activity