Other caches around this code use 500000 entries, so just copy that.

This cache was seeing constant evictions at its former size (70k).

With a size of 200k it now stabilizes at ~122k entries with a
significant drop in server load.

Give more memory to the apps and less to postgres.

Kristian Klausen authored 3 years ago and Jelle van der Waa committed 3 years ago

With the switch to Rspamd from SpamAssassin it isn't faster anymore,
because Rspamd is configured as a global milter where SA was configured
as a smtpd_proxy_filter/content_filter for smtp/submission.

Be explicit about the ports used and enable packet fingerprinting
(checksumming). Both for better compatibility.

Now that login is restricted to staff members.

Closes #94

No need to store secrets for more than one matrix server.

Element uses a statically configured list of homeservers with disabled
presence. Without hosting our own Element this makes for pretty poor UI.

This reverts commit 82ac30f9.

Bump the mem settings higher again, now that we know they're not the
cause. Also increase the maximum connection count.

Less work_mem because this explodes easily. It's per-operation, which a
query can have multiple of, and also across multiple worker threads.

More shared_buffers and effective_cache_size as these are global.

Not sure if this is relevant, but postfix only listens on IPv4. This is
what I just tested as working, so keep it this way.