* disable modsecurity on error page
* fix modsecurity error pages test
* fix variable in nginx template
* disable modsecurity on all internal locations
* fix pipeline checks for gofmt
Signed-off-by: Florian Michel <florianmichel@hotmail.de>
X-CustomHeader looks more like an example than a header we would want to
accept in production. Added Range as a useful header that enables
operations on resources that can be fetched in chunks.
* nginx 1.19.10 keepalive_time parameter
* nginx v1.19.10 base image
* keepalive_time documentation
* base image
* restore base image
* e2e test
* replace default value in test
* Initial work on chrooting nginx process
* More improvements in chroot
* Fix charts and some file locations
* Fix symlink on non chrooted container
* fix psp test
* Add e2e tests to chroot image
* Fix logger
* Add internal logger in controller
* Fix overlay for chrooted tests
* Fix tests
* fix boilerplates
* Fix unittest to point to the right pid
* Fix PR review
* Add keepalive support for auth requests
* Fix typo
* Address PR comments
* Log warning when auth-url contains variable in its host:port
* Generate upstream name without replacing dots to underscores in server name
* Add comment in the nginx template when the keepalive upstream block is referenced
* Workaround for auth_request module ignores keepalive in upstream block
* The `auth_request` module does not support HTTP keepalives in upstream block:
https://trac.nginx.org/nginx/ticket/1579
* As a workaround we use ngx.location.capture but unfortunately it does not
support HTTP/2 so `use-http2` configuration parameter is needed.
* Handle PR comments
* Address PR comments
* Handle invalid values for int parameters
* Handle PR comments
* Fix e2e test
When the ingress controller loads certificates (new ones or following a
secret update), it performs a series of check to ensure its validity.
In our systems, we detected a case where, when the secret object is
compromised, for example when the certificate does not match the secret
key, different pods of the ingress controller are serving a different
version of the certificate.
This behaviour is due to the cache mechanism of the ingress controller,
keeping the last known certificate in case of corruption. When this
happens, old ingress-controller pods will keep serving the old one,
while new pods, by failing to load the corrupted certificates, would
use the default certificate, causing invalid certificates for its
clients.
This generates a random error on the client side, depending on the
actual pod instance it reaches.
In order to allow detecting occurences of those situations, add a metric
to expose, for all ingress controlller pods, detailed informations of
the currently loaded certificate.
This will, for example, allow setting an alert when there is a
certificate discrepency across all ingress controller pods using a query
similar to `sum(nginx_ingress_controller_ssl_certificate_info{host="name.tld"})by(serial_number)`
This also allows to catch other exceptions loading certificates (failing
to load the certificate from the k8s API, ...
Co-authored-by: Daniel Ricart <danielricart@users.noreply.github.com>
Co-authored-by: Daniel Ricart <danielricart@users.noreply.github.com>
* fix inconsistent-label-cardinality
for prometheus metrics: nginx_ingress_controller_requests
* add host to collectorLabels only if metricsPerHost is true
* Disabled default modsecurity_rules_file if modsecurity-snippet is specifed
The default modsecurity_rules_file overwrites the ModSecurity-snippet if it is specified with custom config settings like "SecRuleEngine On". This will not let Modsecurity be in blocking mode even if "SecRuleEngine On" is specified in the ModSecurity-snippet configuration
* Remove unnecessary comments
Only have the default Modsecurity conf settings in case Modsecurity configuration snippet is not present and remove unnecessary comments
* Fixed modsecurity default file only if Modsecurity snippet present
Fixed if condition Modsecurity snippet present have modsecurity default config file
* Added e2e test to disabling modsecurity conf
Added e2e in case modsecurity-snippet enabled to disable settings in default modsecurity.conf
* Validate writing to a different location
Validate also modsecurity to write to a different location instead of the default directory
* Fixed the formatting
* Fixed if empty ModsecuritySnippet
* Fixed ModsecuritySnippet condition
* Fixed the condition also in ingress controller template
* Removed the default config condition in ingress controller template
* Fixed the default config condition in ingress controller template
* Fixed pull-ingress-nginx-test
* Revert "Fixed the default config condition in ingress controller template"
This reverts commit 9d38eca40f.
* Revert template_test
* Adjusted the formating %v
* Add Initial support for multiple cors origins in nginx
- bump cluster version for `make dev-env`
- add buildOriginRegex function in nginx.tmpl
- add e2e 4 e2e tests for cors.go
- refers to feature request #5496
* add tests + use search to identify '*' origin
* add tests + use search to identify '*' origin
Signed-off-by: Christopher Larivière <lariviere.c@gmail.com>
* fix "should enable cors test" looking at improper values
* Modify tests and add some logic for origin validation
- add origin validation in cors ingress annotations
- add extra tests to validate regex
- properly escape regex using "QuoteMeta"
- fix some copy/paste errors
* add TrimSpace and length validation before adding a new origin
* modify documentation for cors and remove dangling comment
* add support for optional port mapping on origin
* support single-level wildcard subdomains + tests
* Remove automatic `*` fonctionality from incorrect origins
- use []string instead of basic string to avoid reparsing in template.go
- fix typo in docs
- modify template to properly enable only if the whole block is enabled
- modify cors parsing
- test properly by validating that the value returned is the proper
origin
- update unit tests and annotation tests
* Re-add `*` when no cors origins are supplied + fix tests
- fix e2e tests to allow for `*`
- re-add `*` to cors parsing if trimmed cors-allow-origin is empty
(supplied but empty) and if it wasn't supplied at all.
* remove unecessary logic for building cors origin + remove comments
- add some edge cases in e2e tests
- rework logic for building cors origin
there was no need for logic in template.go for buildCorsOriginRegex
if there is a `*` it ill be short-circuited by first if.
if it's a wildcard domain or any domain (without a wildcard), it MUST
match the main/cors.go regex format.
if there's a star in a wildcard domain, it must be replaced with
`[A-Za-z0-9]+`
* add missing check in e2e tests
It is possible to change this behavior on an ingress level, which works
well when you only have a few of them. When running several dozen
ingress and with a high change rate of running pods it makes it easier
to define this configuration on a global level.
This change is completely backwards compatible, only adding the
possibility of defining a new key in the configmap.
<!--- Provide a general summary of your changes in the Title above --->
<!--- Why is this change required? What problem does it solve? -->
Introduces the CLI command flag `--disable-full-test`
By default, it doesn't alter the current behavior of the tests performed by the admission controller.
With or Without the flag, a full checkOverlap is actioned, without any alteration
and the object `pcfg` is created with the whole set of ingreses.
If the flag is set to true, it does manipulate the size of `pcfg` up to the content of $this single ingress.
This is achieved by overriding pcfg content by just the last slice that got recently appended to the object `ings`
```
if n.cfg.DisableFullValidationTest {
_, _, pcfg = n.getConfiguration(ings[len(ings)-1:])
}
```
The following steps of generateTemplate and testTemplate are significally reduced to a signle scenario
```
content, err := n.generateTemplate(cfg, *pcfg)
...
err = n.testTemplate(content)
```
This flag doesn't avoid the proper testing of collisions, neither bad syntaxis within the rendered
configuration of the ingress.
But it does eliminate a scenario, which I wasn't able to produce, where by for some reason even proper rendering
and valid values, without collisions of host/path may end into an invalid nginx.conf
The reasoning for this Feature is:
- Test duration increases by the number of ingresses in the cluster.
- File size grows to very important numbers 150-200Mb on clusters with just 2000~ ingresses.
- Tests in that scenario, takes approximately 20s using the last 0.48.1 improvements
- Produces a considerable memory consumption, as well as CPU, compute, that affects directly the containers
that serve traffic.
Since the flag is trully optional, and by default is disabled I fell as a good thing to have that can definitively
help on large-scale scenarios that still want to have a reasonable set of tests in place at a lower cost.
<!--- What types of changes does your code introduce? Put an `x` in all the boxes that apply: -->
- [ ] Bug fix (non-breaking change which fixes an issue)
- [X ] New feature (non-breaking change which adds functionality)
- [ ] Breaking change (fix or feature that would cause existing functionality to change)
<!--- Please describe in detail how you tested your changes. -->
<!--- Include details of your testing environment, and the tests you ran to -->
<!--- see how your change affects other areas of the code, etc. -->
Tested with the build kit the following scenarios on a cluster with 1000~ ingresses:
- With Flag Disabled or Flag, not present (current status as per 0.48.1)
collision scenario (wrong snippet content):
`kubectl apply -f ../collision-syntax.yaml 0.18s user 0.05s system 3% cpu 6.639 total`
collisions scenario (duplicated host):
`kubectl apply -f ../collision-host.yaml 0.17s user 0.05s system 3% cpu 6.245 total`
create/update:
`kubectl apply -f ing-215.yaml 0.16s user 0.05s system 3% cpu 5.845 total`
- With Flag Enabled (true):
collision scenario (wrong snippet content):
`kubectl apply -f ../collision.yaml 0.18s user 0.02s system 57% cpu 0.347 total`
collision scenario (duplicated host):
`kubectl apply -f ../collision.yaml 0.21s user 0.06s system 85% cpu 0.318 total`
create/update:
`kubectl apply -f ing-973.yaml 0.17s user 0.03s system 72% cpu 0.271 total`
As part of the test, I did verified that the created nginx for the test was of a smaller size, and that it didnt affect negatively the final nginx.conf (of a much larger side) where this was merged by the next steps in place after the validation. I couldn't observe any other change in the behaviour and so far the routine looks simple and non harmful.
<!--- Go over all the following points, and put an `x` in all the boxes that apply. -->
<!--- If you're unsure about any of these, don't hesitate to ask. We're here to help! -->
- [x] My change requires a change to the documentation.
- [x] I have updated the documentation accordingly.
- [x] I've read the [CONTRIBUTION](https://github.com/kubernetes/ingress-nginx/blob/main/CONTRIBUTING.md) guide
- [ ] I have added tests to cover my changes.
- [ ] All new and existing tests passed.
For the test part, I would need to understand the placement and test case that this would require, I wasn't able to see an existing scenario for this
Normally Ingress sinchronization for Services is triggered when
corresponding Service's Endpoints are added, deleted or modified.
Services of type ExternalName, however, do not have any endpoints
and hence do not trigger Ingress synchronization as only Update
events are being watched. This commit makes sure that Update and
Delete Service events also enqueue a syncIngress task.
* Change statusSync.runningAddresses() return type
Previously, this method returning a string slice containing the resolved
IP addresses / FQDNs to sync onto the Ingress. It was then converted
just before use into a slice of LoadBalancerIngresses.
This commit changes this logic so that this method generates
LoadBalancerIngress objects directly, and returns these. This has two
main benefits:
- Future work in syncing _both_ hostname and IP, or any other fields
that may be used in future (eg Ports), is now supported.
- There is less need to rely on net.ParseIP() to determine if a value is
an IP address or Hostname, as this can be correctly assigned at
generation time based on where each value came from.
* Sync both IP and Hostname to Ingress Status
Previously, if the IP address was set on a PublishService's
LoadBalancerIngress entries, only that would be synced. Hostname was
only synced as a fallback when the IP address was missing.
Now, both fields are checked independantly and both are synced if
present.
* Fix indentation of nested list in AuthTLS annotations
Also, put `<annotation>`: <description text>` on a single line in
Markdown markup, which will match what gets rendered eventually.
On the other hand, for the line on auth-tls-secret (This annotation
expects the Secret name in the form "namespace/secretName"), its
Markdown markup suggests that the author wanted the line to start on its
own line, but currently this gets rendered on the same line. It's nice
for this to be on its own line, since it's kind of a "note" about the
annotation syntax. Format/indent the markup appropriately so that it
shows up on its line.
* Fix indentation of nested list in CORS annotations
Also, put `<annotation>`: <description text>` on a single line in
Markdown markup, which will match what gets rendered eventually.
On the other hand, for lines noting the allowed characters (This is a
multi-valued field...), its Markdown markup suggests that the author
wanted the line to start on its own line, but currently this gets
rendered on the same line. It's nice for this to be on its own line,
since it's kind of a "note" about the annotation syntax. Format/indent
the markup appropriately so that it shows up on its line.
* Replace f.HTTPTestClientWithTLSConfig() in AuthTLS E2E, the odd one out for requests without client certs
* Demonstrate and document that auth-tls-secret enables the other AuthTLS annotations like verify client, depth
* Split E2E for auth-tls-error-page and *-pass-certificate-to-upstream
