We'd like to start using the DNS load balancer in the latest version of gRPC. That means putting all IPs for a service under a single hostname (or using a SRV record, but we're not taking that path). This change adds an sd-test-srv to act as our service discovery DNS service. It returns both Boulder IP addresses for any A lookup ending in ".boulder". This change also sets up the Docker DNS for our boulder container to defer to sd-test-srv when it doesn't know an answer.
sd-test-srv doesn't know how to resolve public Internet names like `github.com`. Resolving public names is required for the `godep-restore` test phase, so this change breaks out a copy of the boulder container that is used only for `godep-restore`.
This change implements a shim of a DNS resolver for gRPC, so that we can switch to DNS-based load balancing with the currently vendored gRPC, then when we upgrade to the latest gRPC we won't need a simultaneous config update.
Also, this change introduces a check at the end of the integration test that each backend received at least one RPC, ensuring that we are not sending all load to a single backend.
In staging/prod we use `doNotForceCN: false` for both the RA & CA
config. Switching this to `true` is blocked on CABF work that will
likely take considerable time.
In the short-term we should use `doNotForceCN: false` in `test/config`
and only use `doNotForceCN: true` in `test/config-next`.
We're currently stuck on gRPC v1.1 because of a breaking change to certificate validation in gRPC 1.8. Our gRPC balancer uses a static list of multiple hostnames, and expects to validate against those hostnames. However gRPC expects that a service is one hostname, with multiple IP addresses, and validates all those IP addresses against the same hostname. See grpc/grpc-go#2012.
If we follow gRPC's assumptions, we can rip out our custom Balancer and custom TransportCredentials, and will probably have a lower-friction time in general.
This PR is the first step in doing so. In order to satisfy the "multiple IPs, one port" property of gRPC backends in our Docker container infrastructure, we switch to Docker's user-defined networking. This allows us to give the Boulder container multiple IP addresses on different local networks, and gives it different DNS aliases in each network.
In startservers.py, each shard of a service listens on a different DNS alias for that service, and therefore a different IP address. The listening port for each shard of a service is now identical.
This change also updates the gRPC service certificates. Now, each certificate that is used in a gRPC service (as opposed to something that is "only" a client) has three names. For instance, sa1.boulder, sa2.boulder, and sa.boulder (the generic service name). For now, we are validating against the specific hostnames. When we update our gRPC dependency, we will begin validating against the generic service name.
Incidentally, the DNS aliases feature of Docker allows us to get rid of some hackery in entrypoint.sh that inserted entries into /etc/hosts.
Note: Boulder now has a dependency on the DNS aliases feature in Docker. By default, docker-compose run creates a temporary container and doesn't assign any aliases to it. We now need to specify docker-compose run --use-aliases to get the correct behavior. Without --use-aliases, Boulder won't be able to resolve the hostnames it wants to bind to.
When the WFE calls the RA the RA creates a sub context which is cancelled when
the RPC returns. Because we were spawning the publisher RPC calls in a
goroutine with the context from ra.IssueCertificate as soon as
ra.IssueCertificate returned that context was being canceled which in turn
canceled the publisher RPC calls.
Instead of using the RA RPC context simply use a `context.Background()` so
that the RPC context doesn't break these submissions. Also return to
pre-features.CancelCTSubmissions behavior where precert submissions would
be canceled once we retrieved SCTs from the winning logs instead of relying
on the magic behavior of the RA RPC canceling them itself.
Submits final certificates to any configured CT logs. This doesn't introduce a feature flag as it is config gated, any log we want to submit final certificates to needs to have it's log description updated to include the `"submitFinalCerts": true` field.
Fixes#3605.
During periods of peak load, some RPCs are significantly delayed (on the order of seconds) by client-side blocking. HTTP/2 clients have to obey a "max concurrent streams" setting sent by the server. In Go's HTTP/2 implementation, this value [defaults to 250](https://github.com/golang/net/blob/master/http2/server.go#L56), so the gRPC default is also 250. So whenever there are more than 250 requests in progress at a time, additional requests will be delayed until there is a slot available.
During this peak load, we aren't hitting limits on CPU or memory, so we should increase the max concurrent streams limit to take better advantage of our available resources. This PR adds a config field to do that.
Fixes#3641.
gRPC passes deadline information through the RPC boundary, but client and server have the same deadline. Ideally we'd like the server to have a slightly tighter deadline than the client, so if one of the server's onward RPCs or other network calls times out, the server can pass back more detailed information to the client, rather than the client timing out the server and losing the opportunity to log more detailed information about which component caused the timeout.
In this change, I subtract 100ms from the deadline on the server side of our interceptors, using our existing serverInterceptor. I also check that there is at least 100ms remaining in which to do useful work, so the server doesn't begin a potentially expensive task only to abort it.
Fixes#3608.
Adds SCT embedding to the certificate issuance flow. When a issuance is requested a precertificate (the requested certificate but poisoned with the critical CT extension) is issued and submitted to the required CT logs. Once the SCTs for the precertificate have been collected a new certificate is issued with the poison extension replace with a SCT list extension containing the retrieved SCTs.
Fixes#2244, fixes#3492 and fixes#3429.
Right now we check safe browsing at new-authz time, which introduces a possible
external dependency when calling new-authz. This is usually fine, since most safe
browsing checks can be satisfied locally, but when requests have to go external,
it can create variance in new-authz timing.
Fixes#3491.
This code was never enabled in production. Our original intent was to
ship this as part of the ACMEv2 API. Before that could happen flaws were
identified in TLS-SNI-01|02 that resulted in TLS-SNI-02 being removed
from the ACME protocol. We won't ever be enabling this code and so we
might as well remove it.
Add a set of logs which will be submitted to but not relied on for their SCTs,
this allows us to test submissions to a particular log or submit to a log which
is not yet approved by a browser/root program.
Also add a feature which stops cancellations of remaining submissions when racing
to get a SCT from a group of logs.
Additionally add an informational log that always times out in config-next.
Fixes#3464 and fixes#3465.
Removes usage of the `EnforceChallengeDisable` feature, the feature itself is not removed as it is still configured in staging/production, once that is fixed I'll submit another PR removing the actual flag.
This keeps the behavior that when authorizations are retrieved from the SA they have their challenges populated, because that seems to make the most sense to me? It also retains TLS re-validation.
Fixes#3441.
Boulder is fairly noisy about gRPC connection errors. This is a mixed
blessing: Our gRPC configuration will try to reconnect until it hits
an RPC deadline, and most likely eventually succeed. In that case,
we don't consider those to really be errors. However, in cases where
a connection is repeatedly failing, we'd like to see errors in the
logs about connection failure, rather than "deadline exceeded." So
we want to keep logging of gRPC errors.
However, right now we get a lot of these errors logged during
integration tests. They make the output hard to read, and may disguise
more serious errors. So we'd like to avoid causing such errors in
normal integration test operation.
This change reorders the startup of Boulder components by their gRPC
dependencies, so everything's backend is likely to be up and running
before it starts. It also reverses that order for clean shutdowns,
and waits for each process to exit before signalling the next one.
With these changes, I still got connection errors. Taking listenbuddy
out of the gRPC path fixed them. I believe the issue is that
listenbuddy is not a truly transparent proxy. In particular, it
accepts an inbound TCP connection before opening an outbound TCP
connection. If opening that outbound connection results in "connection
refused," it closes the inbound connection. That means gRPC sees a
"connection closed" (or "connection reset"?) rather than "connection
refused". I'm guessing it handles those cases differently, explaining
the different error results.
We've been using listenbuddy to trigger disconnects while Boulder is
running, to ensure that gRPC's reconnect code works. I think we can
probably rely on gRPC's reconnect to work. The initial problem that
led us to start testing this was a configuration problem; now that
we have the configuration we want, we should be fine and don't need
to keep testing reconnects on every integration test run.
For the upcoming SCT embedding changes, it will be useful to have a CT test server that blocks for nontrivial amounts of time before responding. This change introduces a config file for `ct-test-srv` that can be used to set up multiple "personalities" on various ports. Each personality can have a "latencySchedule" that determines how long it will sleep before servicing responding to a submission.
This change also introduces two new "personalities" on :4510 and :4511, plus configures CTLogGroups in the RA. Having four CT log personalities allows us to simulate two nontrivial log groups.
Note: This triggers Publisher to emit audit errors on timed-out submissions. We may want to make Publisher not treat those as errors, and instead only log an error if a whole log group fails.
This change adds a feature flag, TLSSNIRevalidation. When it is enabled, Boulder
will create new authorization objects with TLS-SNI challenges if the requesting
account has issued a certificate with the relevant domain name, and was the most
recent account to do so*. This setting overrides the configured list of
challenges in the PolicyAuthority, so even if TLS-SNI is disabled in general, it
will be enabled for revalidation.
Note that this interacts with EnforceChallengeDisable. Because
EnforceChallengeDisable causes additional checked at validation time and at
issuance time, we need to update those two places as well. We'll send a
follow-up PR with that.
*We chose to make this work only for the most recent account to issue, even if
there were overlapping certificates, because it significantly simplifies the
database access patterns and should work for 95+% of cases.
Note that this change will let an account revalidate and reissue for a domain
even if the previous issuance on that account used http-01 or dns-01. This also
simplifies implementation, and fits within the intent of the mitigation plan: If
someone previously issued for a domain using http-01, we have high confidence
that they are actually the owner, and they are not going to "steal" the domain
from themselves using tls-sni-01.
Also note: This change also doesn't work properly with ReusePendingAuthz: true.
Specifically, if you attempted issuance in the last couple days and failed
because there was no tls-sni challenge, you'll still have an http-01 challenge
lying around, and we'll reuse that; then your client will fail due to lack of
tls-sni challenge again.
This change was joint work between @rolandshoemaker and @jsha.
This updates the PA component to allow authorization challenge types that are globally disabled if the account ID owning the authorization is on a configured whitelist for that challenge type.
This PR implements issuance for wildcard names in the V2 order flow. By policy, pending authorizations for wildcard names only receive a DNS-01 challenge for the base domain. We do not re-use authorizations for the base domain that do not come from a previous wildcard issuance (e.g. a normal authorization for example.com turned valid by way of a DNS-01 challenge will not be reused for a *.example.com order).
The wildcard prefix is stripped off of the authorization identifier value in two places:
When presenting the authorization to the user - ACME forbids having a wildcard character in an authorization identifier.
When performing validation - We validate the base domain name without the *. prefix.
This PR is largely a rewrite/extension of #3231. Instead of using a pseudo-challenge-type (DNS-01-Wildcard) to indicate an authorization & identifier correspond to the base name of a wildcard order name we instead allow the identifier to take the wildcard order name with the *. prefix.
This PR changes the VA's singleDialTimeout value from 5 * time.Second to 10 * time.Second. This will give slower servers a better chance to respond, especially for the multi-VA case where n requests arrive ~simultaneously.
This PR also bumps the RA->VA timeout by 5s and the WFE->RA timeout by 5s to accommodate the increased dial timeout. I put this in a separate commit in case we'd rather deal with this separately.
Makes a couple of changes:
* Change `SubmitToCT` to make submissions to each log in parallel instead of in serial, this prevents a single slow log from eating up the majority of the deadline and causing submissions to other logs to fail
* Remove the 'submissionTimeout' field on the publisher since it is actually bounded by the gRPC timeout as is misleading
* Add a timeout to the CT clients internal HTTP client so that when log servers hang indefinitely we actually do retries instead of just using the entire submission deadline. Currently set at 2.5 minutes
Fixes#3218.
We used to use TCP because we would request DNSSEC records from Unbound, and
they would always cause truncated records when present. Now that we no longer
request those (#2718), we can use UDP. This is better because the TCP serving
paths in Unbound are likely less thoroughly tested, and not optimized for high
load. In particular this may resolve some availability problems we've seen
recently when trying to upgrade to a more recent Unbound.
Note that this only affects the Boulder->Unbound path. The Unbound->upstream
path is already UDP by default (with TCP fallback for truncated ANSWERs).
Fixes#2889.
VA now implements two gRPC services: VA and CAA. These both run on the same port, but this allows implementation of the IsCAAValid RPC to skip using the gRPC wrappers, and makes it easier to potentially separate the service into its own package in the future.
RA.NewCertificate now checks the expiration times of authorizations, and will call out to VA to recheck CAA for those authorizations that were not validated recently enough.
The existing ReusePendingAuthz implementation had some bugs:
It would recycle deactivated authorizations, which then couldn't be fulfilled. (#2840)
Since it was implemented in the SA, it wouldn't get called until after the RA checks the Pending Authorizations rate limit. Which means it wouldn't fulfill its intended purpose of making accounts less likely to get stuck in a Pending Authorizations limited state. (#2831)
This factors out the reuse functionality, which used to be inside an "if" statement in the SA. Now the SA has an explicit GetPendingAuthorization RPC, which gets called from the RA before calling NewPendingAuthorization. This happens to obsolete #2807, by putting the recycling logic for both valid and pending authorizations in the RA.
Adds a basic truncated modulus hash check for RSA keys that can be used to check keys against the Debian `{openssl,openssh,openvpn}-blacklist` lists of weak keys generated during the [Debian weak key incident](https://wiki.debian.org/SSLkeys).
Testing is gated on adding a new configuration key to the WFE, RA, and CA configs which contains the path to a directory which should contain the weak key lists.
Fixes#157.
This removes the config and code to output to statsd.
- Change `cmd.StatsAndLogging` to output a `Scope`, not a `Statter`.
- Remove the prefixing of component name (e.g. "VA") in front of stats; this was stripped by `autoProm` but now no longer needs to be.
- Delete vendored statsd client.
- Delete `MockStatter` (generated by gomock) and `mocks.Statter` (hand generated) in favor of mocking `metrics.Scope`, which is the interface we now use everywhere.
- Remove a few unused methods on `metrics.Scope`, and update its generated mock.
- Refactor `autoProm` and add `autoRegisterer`, which can be included in a `metrics.Scope`, avoiding global state. `autoProm` now registers everything with the `prometheus.Registerer` it is given.
- Change va_test.go's `setup()` to not return a stats object; instead the individual tests that care about stats override `va.stats` directly.
Fixes#2639, #2733.
Prior to this PR if a domain was an exact match to a public suffix
list entry the certificates per name rate limit was applied based on the
count of certificates issued for that exact name and all of its
subdomains.
This PR introduces an exception such that exact public suffix
matches correctly have the certificate per name rate limit applied based
on only exact name matches.
In order to accomplish this a new RPC is added to the SA
`CountCertificatesByExactNames`. This operates similar to the existing
`CountCertificatesByNames` but does *not* include subdomains in the
count, only exact matches to the names provided. The usage of this new
RPC is feature flag gated behind the "CountCertificatesExact" feature flag.
The RA unit tests are updated to test the new code paths both with and
without the feature flag enabled.
Resolves#2681
Deletes github.com/streadway/amqp and the various RabbitMQ setup tools etc. Changes how listenbuddy is used to proxy all of the gRPC client -> server connections so we test reconnection logic.
+49 -8,221 😁Fixes#2640 and #2562.
I think these are all the necessary changes to implement TLS-SNI-02 validations, according to the section 7.3 of draft 05:
https://tools.ietf.org/html/draft-ietf-acme-acme-05#section-7.3
I don't have much experience with this code, I'll really appreciate your feedback.
Signed-off-by: David Calavera <david.calavera@gmail.com>
Set authorizationLifetimeDays to 60 across both config and config-next.
Set NumSessions to 2 in both config and config-next. A decrease from 10 because pkcs11-proxy (or pkcs11-daemon?) seems to error out under load if you have more sessions than CPUs.
Reorder parallelGenerateOCSPRequests to match config-next.
Remove extra tags for parsing yaml in config objects.
Previously, a given binary would have three TLS config fields (CA cert, cert,
key) for its gRPC server, plus each of its configured gRPC clients. In typical
use, we expect all three of those to be the same across both servers and clients
within a given binary.
This change reuses the TLSConfig type already defined for use with AMQP, adds a
Load() convenience function that turns it into a *tls.Config, and configures it
for use with all of the binaries. This should make configuration easier and more
robust, since it more closely matches usage.
This change preserves temporary backwards-compatibility for the
ocsp-updater->publisher RPCs, since those are the only instances of gRPC
currently enabled in production.
Previously, all gRPC services used the same client and server certificates. Now,
each service has its own certificate, which it uses for both client and server
authentication, more closely simulating production.
This also adds aliases for each of the relevant hostnames in /etc/hosts. There
may be some issues if Docker decides to rewrite /etc/hosts while Boulder is
running, but this seems to work for now.
- Remove spinner from test.js. It made Travis logs hard to read.
- Listen on all interfaces for debugAddr. This makes it possible to check
Prometheus metrics for instances running in a Docker container.
- Standardize DNS timeouts on 1s and 3 retries across all configs. This ensures
DNS completes within the relevant RPC timeouts.
- Remove RA service queue from VA, since VA no longer uses the callback to RA on
completing a challenge.
Adds a gRPC server to the SA and SA gRPC Clients to the WFE, RA, CA, Publisher, OCSP updater, orphan finder, admin revoker, and expiration mailer.
Also adds a CA gRPC client to the OCSP Updater which was missed in #2193.
Fixes#2347.
With the current gRPC design the CA talks directly to the Publisher when calling SubmitToCT which crosses security bounadries (secure internal segment -> internet facing segment) which is dangerous if (however unlikely) the Publisher is compromised and there is a gRPC exploit that allows memory corruption on the caller end of a RPC which could expose sensitive information or cause arbitrary issuance.
Instead we move the RPC call to the RA which is in a less sensitive network segment. Switching the call site from the CA -> RA is gated on adding the gRPC PublisherService object to the RA config.
Fixes#2202.
Jacob Hoffman-Andrews