trafficserver-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "hzwulibin"<>
Subject Re: Re: Re: What's the common usecase of TrafficServer Cluster
Date Fri, 31 Jul 2015 11:15:15 GMT
Hi,  yongming

I see. Means we can't avoid another software here.
Sure, thank you very much!



发件人:Yongming Zhao <>
发送时间:2015-07-31 19:07
主题:Re: Re: What's the common usecase of TrafficServer Cluster

The Cluster is a Cache Cluster, which is a cache storage expansion to all the cluster members,
that means, the cluster should only be consider as a single caching host with multiple IP
interfaces. no high available service as in lvs and others in the cluster. In the Cluster
type = 1, due to the consistent hashing used in the content storing, the URL store is load
balanced to all the cluster members.

in production, you may need to add high available service such as GSLB and LVS etc, or do
the balance/failover your own way, choose on you own needs.

back in the inktomi age, the WCCP can work with Cluster to provide a more efficient management
in local caching deployment, but it may not work that well now.

- Yongming Zhao 赵永明

在 2015年7月29日,上午8:42,hzwulibin <> 写道:

In my case, i want to use trafficserver as a cache proxy; so my application is
configured to get data from the server A if no cluster situation. But i'm
wondering if it will still works well in the cluster.

You memtioned load-bancing, so does it need a load bance between
my application and trafficserver cluster? Or should i deal with it in my

Yeah, maybe both are fine, anyway, i want my model be as simple as
possible. Any suggestions about this?




发件人:Reindl Harald <>
发送时间:2015-07-28 19:34
主题:Re: What's the common usecase of TrafficServer Cluster

Am 28.07.2015 um 13:22 schrieb hzwulibin: 
> Hi, everyone 
> Today i setup the TrafficServer cluster. However, i have some questions. 
> How the cluster supply the service to the client, do we need other 
> solftware like nginx or haproxy? 
> Just want to know the most common usecase! 
> In my thinking now, for example, there are three nodes in the clustre A, 
> B, C; 
> Some clients configed to use A,  when A is down, those clients lost the 
> service. 
> So the clients should config to use all ip of A, B, C, or need a medial 
> solftware like nginx? 
> Can anyone say some common usecase here? 

i don't get "Some clients configed to use A" 

normally you have an A record returning 3 IP's for load-balancing and  
nginx don't make sense in the picture because it plays on the same level  
as ATS itself: a proxy 
View raw message