make initdb slows to a crawl/fails when docker-compose is compose v2

[woodruffw]

Noticed this while trying to stand up a local instance of Warehouse with compose v2 enabled:

Observed behavior:

1. Running make initdb produces much more output on the terminal than it normally does

2. The console scrollback for the db service is filled with lots of SQL syntax errors, e.g.:

   warehouse-db-1             | 2022-02-17 17:02:49.596 UTC [35] ERROR:  syntax error at or near "0.0" at character 1
   

I let make initdb run for about ~5 minutes, which is several times longer than it takes to complete when using compose v1, without visible progress beyond the initial SQL load phase.

Reproduction steps:

These steps are for macOS.

1. Open up the Docker for Mac preferences pane (Cmd + ,)
2. Go to the "General" page
3. Click the checkbox for "Use Docker Compose V2" (screenshot below)
4. Restart the Docker daemon
5. make serve
6. make initdb (in a separate window)

[di]

Are you able to narrow down which of the make initdb commands is the source of the error?

https://github.com/pypa/warehouse/blob/968f11fe99b2e0eba6c833250dc1e007a208cd33/Makefile#L142-L149

I'm guessing it's probably this one:

https://github.com/pypa/warehouse/blob/968f11fe99b2e0eba6c833250dc1e007a208cd33/Makefile#L146

[di]

Parts of #9993 might be a potential solution for this.

[woodruffw]

Yep, it's that one. My guess is that compose V2 has broken/buggy pipe behavior, since that's been a consistent problem across docker and docker-compose over the years.

[woodruffw]

Parts of #9993 might be a potential solution for this.

I can give that changeset a try in a moment!

[miketheman]

If the pipe behavior is the culprit, did this fix it? #10978

[woodruffw]

If the pipe behavior is the culprit, did this fix it? #10978

I'll re-try this afternoon and find out.

[di]

@woodruffw Have you been able to determine if this is resolved?

[woodruffw]

Sorry, this fell off my radar! Trying now (as soon as Docker Desktop finishes updating...)

[woodruffw]

It appears to be working now! I see this in my logs:

warehouse-db-1             | 2022-04-13 22:40:19.421 UTC [39] FATAL:  terminating connection due to administrator command
warehouse-db-1             | 2022-04-13 22:40:19.421 UTC [40] FATAL:  terminating connection due to administrator command
warehouse-db-1             | 2022-04-13 22:40:34.177 UTC [60] FATAL:  role "root" does not exist

...but Warehouse does appear to come up correctly and the commands all still exit successfully. Closing.

[miketheman]

For that last message, I've supplied #11182 to fix it.

[woodruffw]

Awesome. I can test that PR tonight. Sent from mobile. Please excuse my brevity.

…

On Apr 14, 2022, at 4:36 PM, Mike Fiedler ***@***.***> wrote:  For that last message, I've supplied #11182 to fix it. — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you modified the open/close state.