You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@mesos.apache.org by William Morgan <wi...@buoyant.io> on 2016/05/04 21:23:40 UTC
interested in a service load balancer / proxy?
Dear Mesos friends,
We have an open source load balancer / proxy called linkerd
<https://linkerd.io> that might be of interest to some of you. Similar to
marathon-lb, it's a request-level load balancer that speaks to service
discovery. Unlike marathon-lb, it doesn't require process restarts. It's
built on Finagle, Twitter's high-volume RPC library, which means it also
can do intelligent things with slow and failing servers, connection
pooling, etc.
It features pluggable service discovery backends (Zookeeper, Marathon,
Consul, etcd), and pluggable protocol support (HTTP, thrift and mux).
It's designed as a sort of out-of-process client-side network stack for
high-volume service communication, but you can also use it as a debug proxy
or an edge traffic server.
If that sounds interesting, check out linkerd.io for more details, or join
us on our Slack channel <http://slack.linkerd.io>. We'll be presenting it
at ApacheCon <http://sched.co/6OJA> and at MesosCon <http://sched.co/6jta>
later this month. In the mean time, here's a screenshot to whet your
appetite:
Hit us up with any questions!
-William
Re: interested in a service load balancer / proxy?
Posted by William Morgan <wi...@buoyant.io>.
Right now it's focused on stateless services, so Cassandra is not a good
starting point. Generally speaking, it will talk to service discovery to
resolve the addresses of a service, and load balance across the addresses
based on response latency, error rates, etc.
(It's a bit more complicated than that, of course. More details here:
https://linkerd.io/doc/0.4.0/overview/#basic-concepts)
Things like blue-green deploys, and ad-hoc canarying and staging of new
(stateless) services, are what we're focused on now.
Is that helpful?
-William
On Thu, May 5, 2016 at 12:21 AM, vincent gromakowski <
vincent.gromakowski@gmail.com> wrote:
> Hi,
> How does it deal with topology discovery between frameworks ? for example
> when connecting from spark to cassandra, the cassandra seed node reponse
> includes all nodes IP and you don't have much control over this mechanism...
>
> 2016-05-04 23:23 GMT+02:00 William Morgan <wi...@buoyant.io>:
>
>> Dear Mesos friends,
>>
>> We have an open source load balancer / proxy called linkerd
>> <https://linkerd.io> that might be of interest to some of you. Similar
>> to marathon-lb, it's a request-level load balancer that speaks to service
>> discovery. Unlike marathon-lb, it doesn't require process restarts. It's
>> built on Finagle, Twitter's high-volume RPC library, which means it also
>> can do intelligent things with slow and failing servers, connection
>> pooling, etc.
>>
>> It features pluggable service discovery backends (Zookeeper, Marathon,
>> Consul, etcd), and pluggable protocol support (HTTP, thrift and mux).
>>
>> It's designed as a sort of out-of-process client-side network stack for
>> high-volume service communication, but you can also use it as a debug proxy
>> or an edge traffic server.
>>
>> If that sounds interesting, check out linkerd.io for more details, or
>> join us on our Slack channel <http://slack.linkerd.io>. We'll be
>> presenting it at ApacheCon <http://sched.co/6OJA> and at MesosCon
>> <http://sched.co/6jta> later this month. In the mean time, here's a
>> screenshot to whet your appetite:
>>
>>
>>
>> Hit us up with any questions!
>>
>> -William
>>
>
>
Re: interested in a service load balancer / proxy?
Posted by vincent gromakowski <vi...@gmail.com>.
Hi,
How does it deal with topology discovery between frameworks ? for example
when connecting from spark to cassandra, the cassandra seed node reponse
includes all nodes IP and you don't have much control over this mechanism...
2016-05-04 23:23 GMT+02:00 William Morgan <wi...@buoyant.io>:
> Dear Mesos friends,
>
> We have an open source load balancer / proxy called linkerd
> <https://linkerd.io> that might be of interest to some of you. Similar to
> marathon-lb, it's a request-level load balancer that speaks to service
> discovery. Unlike marathon-lb, it doesn't require process restarts. It's
> built on Finagle, Twitter's high-volume RPC library, which means it also
> can do intelligent things with slow and failing servers, connection
> pooling, etc.
>
> It features pluggable service discovery backends (Zookeeper, Marathon,
> Consul, etcd), and pluggable protocol support (HTTP, thrift and mux).
>
> It's designed as a sort of out-of-process client-side network stack for
> high-volume service communication, but you can also use it as a debug proxy
> or an edge traffic server.
>
> If that sounds interesting, check out linkerd.io for more details, or
> join us on our Slack channel <http://slack.linkerd.io>. We'll be
> presenting it at ApacheCon <http://sched.co/6OJA> and at MesosCon
> <http://sched.co/6jta> later this month. In the mean time, here's a
> screenshot to whet your appetite:
>
>
>
> Hit us up with any questions!
>
> -William
>