fixes#232505
Implements the new option `security.acme.maxConcurrentRenewals` to limit
the number of certificate generation (or renewal) jobs that can run in
parallel. This avoids overloading the system resources with many
certificates or running into acme registry rate limits and network
timeouts.
Architecture considerations:
- simplicity, lightweight: Concerns have been voiced about making this
already rather complex module even more convoluted. Additionally,
locking solutions shall not significantly increase performance and
footprint of individual job runs.
To accomodate these concerns, this solution is implemented purely in
Nix, bash, and using the light-weight `flock` util. To reduce
complexity, jobs are already assigned their lockfile slot at system
build time instead of dynamic locking and retrying. This comes at the
cost of not always maxing out the permitted concurrency at runtime.
- no stale locks: Limiting concurrency via locking mechanism is usually
approached with semaphores. Unfortunately, both SysV as well as
POSIX-Semaphores are *not* released when the process currently locking
them is SIGKILLed. This poses the danger of stale locks staying around
and certificate renewal being blocked from running altogether.
`flock` locks though are released when the process holding the file
descriptor of the lock file is KILLed or terminated.
- lockfile generation: Lock files could either be created at build time
in the Nix store or at script runtime in a idempotent manner.
While the latter would be simpler to achieve, we might exceed the number
of permitted concurrent runs during a system switch: Already running
jobs are still locked on the existing lock files, while jobs started
after the system switch will acquire locks on freshly created files,
not being blocked by the still running services.
For this reason, locks are generated and managed at runtime in the
shared state directory `/var/lib/locks/`.
nixos/security/acme: move locks to /run
also, move over permission and directory management to systemd-tmpfiles
nixos/security/acme: fix some linter remarks in my code
there are some remarks left for existing code, not touching that
nixos/security/acme: redesign script locking flow
- get rid of subshell
- provide function for wrapping scripts in a locked environment
nixos/acme: improve visibility of blocking on locks
nixos/acme: add smoke test for concurrency limitation
heavily inspired by m1cr0man
nixos/acme: release notes entry on new concurrency limits
nixos/acme: cleanup, clarifications
Fixes#190493
Check if an actual key file exists. This does not
completely cover the work accountHash does to ensure
that a new account is registered when account
related options are changed.
Fixes#191794
Lego threw a permission denied error binding to port 80.
AmbientCapabilities with CAP_NET_BIND_SERVICE was required.
Also added a test for this.
In the process I also found that the CapabilityBoundingSet
was restricting the service from listening on port 80, and
the AmbientCapabilities was ineffective. Fixed appropriately.
Allows configuring many default settings for certificates,
all of which can still be overridden on a per-cert basis.
Some options have been moved into .defaults from security.acme,
namely email, server, validMinDays and renewInterval. These
changes will not break existing configurations thanks to
mkChangedOptionModule.
With this, it is also now possible to configure DNS-01 with
web servers whose virtualHosts utilise enableACME. The only
requirement is you set `acmeRoot = null` for each vhost.
The test suite has been revamped to cover these additions
and also to generally make it easier to maintain. Test config
for apache and nginx has been fully standardised, and it
is now much easier to add a new web server if it follows
the same configuration patterns as those two. I have also
optimised the use of switch-to-configuration which should
speed up testing.
Closes#129838
It is possible for the CA to revoke a cert that has not yet
expired. We must run lego to validate this before expiration,
but we must still ignore failures on unexpired certs to retain
compatibility with #85794
Also changed domainHash logic such that a renewal will only
be attempted at all if domains are unchanged, and do a full
run otherwises. Resolves#147540 but will be partially
reverted when go-acme/lego#1532 is resolved + available.
Commit 3a2e0c36e7 has removed
`--reuse-key` from default renew options, yet the tests still expected
keys not to change. This assertion is now removed, as they are supposed
to change on each renew/change.
As per #121293, I ensured the UMask is set correctly
and removed any unnecessary chmod/chown/chgrp commands.
The test suite already partially covered permissions
checking but I added an extra check for the selfsigned
cert permissions.
With the UMask set to 0023, the
mkdir -p command which creates the webroot
could end up unreadable if the web server
changes, as surfaced by the test suite in #114751
On top of this, the following commands
to chown the webroot + subdirectories was
mostly unnecessary. I stripped it back to
only fix the deepest part of the directory,
resolving #115976, and reintroduced a
human readable error message.
Closes#106565
When generating multiple certificates which all
share the same server + email, lego will attempt
to create an account multiple times. By adding an
account creation target certificates which share
an account will wait for one service (chosen at
config build time) to complete first.
Attempting to reuse keys on a basis different to the cert (AKA,
storing the key in a directory with a hashed name different to
the cert it is associated with) was ineffective since when
"lego run" is used it will ALWAYS generate a new key. This causes
issues when you revert changes since your "reused" key will not
be the one associated with the old cert. As such, I tore out the
whole keyDir implementation.
As for the race condition, checking the mtime of the cert file
was not sufficient to detect changes. In testing, selfsigned
and full certs could be generated/installed within 1 second of
each other. cmp is now used instead.
Also, I removed the nginx/httpd reload waiters in favour of
simple retry logic for the curl-based tests
Testing of certs failed randomly when the web server was still
returning old certs even after the reload was "complete". This was
because the reload commands send process signals and do not wait
for the worker processes to restart. This commit adds log watchers
which wait for the worker processes to be restarted.
- Use an acme user and group, allow group override only
- Use hashes to determine when certs actually need to regenerate
- Avoid running lego more than necessary
- Harden permissions
- Support "systemctl clean" for cert regeneration
- Support reuse of keys between some configuration changes
- Permissions fix services solves for previously root owned certs
- Add a note about multiple account creation and emails
- Migrate extraDomains to a list
- Deprecate user option
- Use minica for self-signed certs
- Rewrite all tests
I thought of a few more cases where things may go wrong,
and added tests to cover them. In particular, the web server
reload services were depending on the target - which stays alive,
meaning that the renewal timer wouldn't be triggering a reload
and old certs would stay on the web servers.
I encountered some problems ensuring that the reload took place
without accidently triggering it as part of the test. The sync
commands I added ended up being essential and I'm not sure why,
it seems like either node.succeed ends too early or there's an
oddity of the vm's filesystem I'm not aware of.
- Fix duplicate systemd rules on reload services
Since useACMEHost is not unique to every vhost, if one cert
was reused many times it would create duplicate entries in
${server}-config-reload.service for wants, before and
ConditionPathExists
Reads a bit more naturally, and now the changes to the
acme-${cert}.service actually reflect what would be needed were you to
do the same in production.
e.g. "for dns-01, your service that needs the cert needs to pull in the
cert"
Shimming out the Let's Encrypt domain name to reuse client configuration
doesn't work properly (Pebble uses different endpoint URL formats), is
recommended against by upstream,[1] and is unnecessary now that the ACME
module supports specifying an ACME server. This commit changes the tests
to use the domain name acme.test instead, and renames the letsencrypt
node to acme to reflect that it has nothing to do with the ACME server
that Let's Encrypt runs. The imports are renamed for clarity:
* nixos/tests/common/{letsencrypt => acme}/{common.nix => client}
* nixos/tests/common/{letsencrypt => acme}/{default.nix => server}
The test's other domain names are also adjusted to use *.test for
consistency (and to avoid misuse of non-reserved domain names such
as standalone.com).
[1] https://github.com/letsencrypt/pebble/issues/283#issuecomment-545123242
Co-authored-by: Yegor Timoshenko <yegortimoshenko@riseup.net>
The recent custom endpoint addition allows us to directly point
certbot to the custom Pebble directory endpoint.
Thanks to that, we can ditch the Pebble patch we were using so far;
making this test maintenance easier.
Let's encrypt bumped ACME to V2. We need to update our nixos test to
be compatible with this new protocol version.
We decided to drop the Boulder ACME server in favor of the more
integration test friendly Pebble.
- overriding cacert not necessary
- this avoids rebuilding lots of packages needlessly
- nixos/tests/acme: use pebble's ca for client tests
- pebble always generates its own ca which has to be fetched
TODO: write proper commit msg :)
* nixos/acme: Fix ordering of cert requests
When subsequent certificates would be added, they would
not wake up nginx correctly due to target units only being triggered
once. We now added more fine-grained systemd dependencies to make sure
nginx always is aware of new certificates and doesn't restart too early
resulting in a crash.
Furthermore, the acme module has been refactored. Mostly to get
rid of the deprecated PermissionStartOnly systemd options which were
deprecated. Below is a summary of changes made.
* Use SERVICE_RESULT to determine status
This was added in systemd v232. we don't have to keep track
of the EXITCODE ourselves anymore.
* Add regression test for requesting mutliple domains
* Deprecate 'directory' option
We now use systemd's StateDirectory option to manage
create and permissions of the acme state directory.
* The webroot is created using a systemd.tmpfiles.rules rule
instead of the preStart script.
* Depend on certs directly
By getting rid of the target units, we make sure ordering
is correct in the case that you add new certs after already
having deployed some.
Reason it broke before: acme-certificates.target would
be in active state, and if you then add a new cert, it
would still be active and hence nginx would restart
without even requesting a new cert. Not good! We
make the dependencies more fine-grained now. this should fix that
* Remove activationDelay option
It complicated the code a lot, and is rather arbitrary. What if
your activation script takes more than activationDelay seconds?
Instead, one should use systemd dependencies to make sure some
action happens before setting the certificate live.
e.g. If you want to wait until your cert is published in DNS DANE /
TLSA, you could create a unit that blocks until it appears in DNS:
```
RequiredBy=acme-${cert}.service
After=acme-${cert}.service
ExecStart=publish-wait-for-dns-script
```
