Re: BUG #18493: COPY FROM STDIN BINARY failure

Hi,

On 2024-06-03 14:47:16 +0000, PG Bug reporting form wrote:
> The following bug has been logged on the website:
>
> Bug reference: 18493
> Logged by: Paul Tobey
> Email address: ptobey(at)rainbird(dot)com
> PostgreSQL version: 16.1

Note that there have been a few minor versions released since 16.1

> There are no foreign keys in the table and all indices are disabled.

What precisely do you mean by that?

Could you show the table definition?

> At the failure point, this is the state indicated in pgAdmin 4. There seems
> to be one IPC:BufferIO for each client with similar content each in the same
> "waiting" state:

Hm. Are there any related processes (i.e. other COPYs, checkpointer,
background writer or autovacuum worker) that are *not* showing this wait
event?

Basically, IPC:BufferIO indicates that another backend is currently performing
IO on that buffer. One likely explanation of the problem is that that other
backend is currently stuck waiting for IO complete, e.g. because of storage
issues.

Does the issue reproduce if you retry the data load?

If you have development tools available, it'd be helpful to get backtraces for
some of the stuck processes.

> (See previous pgsql-bugs mailing list post which includes attachments)

For reference, that's at
https://postgr.es/m/IA0PR11MB7935925E44D2A1E814617EF7A9F32%40IA0PR11MB7935.namprd11.prod.outlook.com

Greetings,

Andres Freund