Race condition during master startup about postdock

[paunin]

Fair point....should be covered - but never faced it on live/test env...hmm, any thoughts how to avoid, may be? Worst option is to add some sleep after wait_db here, but may be there is more elegant way.

[Nemo157]

Maybe there's some way to detect if the database is running via pg_ctl instead of directly.

Or according to this line the initial database start doesn't listen on the pods external IP, so might be possible to switch the host entry to use that to connect to itself instead of using the loopback interface.

[paunin]

I've tried to kill this local loopback in branch no-loopback and you can try it out with tag :no-loopback for postgres image... I will test it ASAP but you can try as well. Let me know if any issue with it.

--

UP: so I did tests...looks fine from my side, but could you test it as well. Thanks.
@Nemo157

[paunin]

#37

[Nemo157]

Yep, that worked fine. I even saw the attempt to check the database right as the initialization happened and it failed as it should.

[paunin]

Cool. merged.