Source: https://whetstone.privatestorage.io/privatestorage/PrivateStorageio/-/merge_requests/237#note_18712

Thanks to @jcalderone for the suggestion!

Fixes privatestorageops#408

This is disallowed in nixos 21.11 as it is a potential security issue.

Take what I learned from @tomprince in
https://whetstone.privatestorage.io/privatestorage/PrivateStorageio/-/merge_requests/220#note_17605 :
Assigning a service a `startAt' will make timer.target start it,
no need to also have it `wantedBy' multiuser.target.

Also add megacli2prom to the path even if it should be already because
of megacli2prom.  Better be safe than sorry.

mkfs silently truncates "zkapissuer-database" to "zkapissuer-datab" and
everything falls apart.

The point is this is where PaymentServer's persistent state goes.  That
includes vouchers but might include more stuff too.

The hardware modules can now configure this fileSystem in a way appropriate
for themselves and the issuer module can enable that configuration when it is
enabled itself.

now all nodes load the issuer module so we can always supply issuer
configuration, where it makes sense to do so, even on systems where the issuer
is not going to be enabled.

if that sounds silly read the comment in default.nix

use the new one we're setting up on a new filesystem

Thanks go out to @tomprince!

Tryfix the starting of the timer before tahoe is updated.

This adds a detailed dashboard for tahoe-lafs running on one node.

Thanks to @tomprince for the suggestion.

Copy everything from how the issuer does it.

literalExample is deprecated and overused. See:
https://whetstone.privatestorage.io/privatestorage/PrivateStorageio/-/merge_requests/201#note_16966

Thanks for the suggestion @jcalderone!

Response time, probe fails and TLS expiry incl. alerts.

The Blackbox exporter can be used to check whether some services do answer;
We'll use it for our HTTPS endpoints.
Especially handy is it's checking for TLS cert expiry.

This change reflects my intention of having an alert fire when a node
is not scrapable for five minutes or longer.  I hadn't understood the
default avg() function well enough before, yielding an alert that would
fire too soon (i.e. any time a node was down, even only one minute for a
reboot).