We have online storage application, where user can upload/download his data from the private space allocated to user similar to Google Drive and Dropbox.
We have micro services like Upload and download. Upload micro service generates unique file identifier for each content uploaded by the user and these content accessible from those unique URLs.
So for each download and upload with got unique URLs like mentioned below
(removed https:// from above mentioned urls due that its was not allowing me tor create topic)
We are using sidecar linkerd setup, and Namerd is separated from linkerd. We have got our all dTabs in consul KV. Namerd fetches all the dTabs from consul and linkerd fetches dTabs from namerd.
So according to the implementation linkerd should not send request to namerd for each and every request for fetching updated dtabs. But in our case as all the request are unique so it’s searching for dtabs entries in namerd
Like i am downloading object mentioned on [2.] first time my request will to till namerd to fetch the dtab, and for next occurrence it will fullfill from linkerd itself. But if user has multiple objects downloading all retests will go till namerd as all object endpoints are unique.
dTabs which i am using for both the micro service mentioned below
/http/*/*/*/upload/* => /#/io.l5d.consul/myapp/upload
/http/*/*/*/download/* => /#/io.l5d.consul/myapp/download
Also attaching, linkerd and namerd configuration. Can you please suggest us in such scenario what are the changes in dTabs or configuration is required so we can optimize our network traffic.
PS: i have replaced the application name from myapp.com in configuration.