You are right, it is too much work to do. For the moment application_name
will suffice!
Best regards, Cristi Boboc
On Friday, November 5, 2021, 06:15:53 PM GMT+2, Vijaykumar Jain
<[email protected]> wrote:
On Fri, Nov 5, 2021, 7:46 PM Michael Lewis <[email protected]> wrote:
For my purposes, app name isn't long enough so we put a comment at the start of
every SQL that has a unique ID generated in the application. This ensures that
we can tell one connection apart from another even when both are coming from
the same feature/action (applicationName is set to this) even using connection
pooling. Before, we had different requests reusing the same pid of course and
if they happened to have the same appName, we didn't know it was a new process
on app side.
While researching options, I did see crunchy proxy using annotations to split
read/write. But I was not sure if any of these options fit well with multiple
toolings , drivers etc.
https://github.com/CrunchyData/crunchy-proxy/blob/master/docs/crunchy-proxy-user-guide.asciidoc#annotations
The other option washttps://github.com/wgliang/pgproxy
or writing your custom plug-in for envoyproxy filter for postgresql
https://www.envoyproxy.io/docs/envoy/latest/intro/arch_overview/other_protocols/postgres
So basically it was adding one more layer in between the already connection
pooling layers via pgbouncer.The biggest concern is when we use ssl, these
session params cannot be read unless we terminate ssl at the proxy, and then
decode, so I gave up :)
