Re: "out of relcache_callback_list slots" after multiple calls to pg_logical_slot_get_binary_changes

Good catch!

At Sun, 19 Feb 2023 02:40:31 +0000, "shiy(dot)fnst(at)fujitsu(dot)com" <shiy(dot)fnst(at)fujitsu(dot)com> wrote in
> init_rel_sync_cache()), but they are not unregistered when it shutdowns. So,
> after multiple calls to the function, MAX_RELCACHE_CALLBACKS is exceeded. This
> is mentioned in the following comment.
>
> /*
> * We can get here if the plugin was used in SQL interface as the
> * RelSchemaSyncCache is destroyed when the decoding finishes, but there
> * is no way to unregister the relcache invalidation callback.
> */
> if (RelationSyncCache == NULL)
> return;
>
> Could we fix it by adding two new function to unregister relcache callback and
> syscache callback? I tried to do so in the attached patch.

I'm pretty sure that everytime an output plugin is initialized on a
process, it installs the same set of syscache/relcache callbacks each
time. Do you think we could simply stop duplicate registration of
those callbacks by using a static boolean? It would be far simpler.

regards.

--
Kyotaro Horiguchi
NTT Open Source Software Center