Hi, On Fri, Jan 23, 2026 at 1:31 PM Michael Paquier <[email protected]> wrote: > > On Fri, Jan 23, 2026 at 12:07:36PM +0800, Xuneng Zhou wrote: > > Thanks Michael — agreed. Patch v7 dropped the extra fast‑exit path and > > kept the change minimal: only switch to STREAMING if the state is > > still CONNECTING, otherwise leave it unchanged. > > Thanks for the newer version of the patch. I have been playing with > it today, even before you have sent this v7, with a set of injection > points to make the WAL receiver wait at some of its steps, then > double-checked that the startup process was able to control the WAL > receiver as it should, flipping primary_conninfo with reloads. AFAIK > as I can see, that felt OK, so applied as a36164e7465f. > -- > Michael
Thanks for checking and applying it. I’m playing with exposing XLogRecoveryCtlData metrics at the SQL level, following your input. I’ll post the patches and possibly start a new thread for discussion. -- Best, Xuneng
