Fix the delegation handling in the *file* and *dnssec* middleware.
Refactor tests a bit and show that they are failling.
Add a Tree printer, cleanups and tests.
Fix wildcard test - should get no answer from empty-non-terminal
Disable this test (for now), it fails weirdly on travis
(and not locally). Initial suspicion that another server
is still reporting (or something). Anyway hard to replicate
locally - disable it for now.
* Fix k8s client to use client-go
* Fix Kubernetes Build Issue
The client-go code requires you to vendor. I have done a hack here
in the Makefile to vendor it to version 1.5. But looks like we
will need to do this the 'right' way soon.
* Convert v1 to api Objects in List Functions
Also removed the endpoint controller which was not used for anything.
The Watch functions may still need the same treatment.
* Vendor client-go release-1.5
* Fix basic SRV feature
This is actually not serving SRV records correctly, but this should
get it to work as it did prior to the k8s client changes. Another
fix will be needed to serve SRV records as defined in the spec.
* Add additional output in test result
Add the response to the test output.
* Fix erroneous test data
* Add ServiceBackend interface
This adds a ServiceBackend interface that is shared between etcd/etcd3
(later) and kubernetes, leading to a massive reduction in code. When
returning the specific records from their backend.
Fixes#273
ListenAddr is the address where the prometheus metric are exported.
This can be used in tests to listen on "localhost:0" and then later
retrieve the metrics from there. It makes the tests indepent on each
other.
When a file is moved into position we should also reload the zones'
content.
This also fixes deadlock bug in the locking, a reload would block any
further lookups.
Add the rcode to the cached item and use this when we synthesize the
answer again. We could also infer the rcode from the reassembled
message, but this seems easier and is only an integer.
Also set the autoritative bit to 0 for all from-cache answers.
Fixes 357
* middleware/metrics: add more metrics
middleware/cache:
Add metrics for number of elements in the cache. Also export the total
size. Update README to detail the new metrics.
middleware/metrics
Move metrics into subpackage called "vars". This breaks the import
cycle and is cleaner. This allows vars.Report to be used in the
the dnsserver to log refused queries.
middleware/metrics: tests
Add tests to the metrics framework. The metrics/test subpackage allows
scraping of the local server. Do a few test scrape of the metrics that
are defined in the metrics middleware.
This also allows metrics integration tests to check if the caching and
dnssec middleware export their metrics correctly.
* update README
* typos
* fix tests
This fix uses docker container for kubectl. Since Kubernetes docker
image hyperkube has already been downloaded and it consists of
kubectl, there is really no need to download kubectl binary again.
This fix cleans up the Kubernetes related travis setup and removes
unneeded scripts.
This fix also fixes several mismatches of the Kubernetes version used,
so that any changes in version in the future only need to update .travis.yml.
Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
The DS record is handled specially in the server ServeDNS mux, but there
was no code that actually called the correct middleware handler chain
when encountering a DS.
This PR fixes that behavoir, additonal bugs has been files to look into
how we are handling delegation (secure and non-secure ones).
This fix adds a `cache_capacity` option to dnssec middleware, so that
it is possible to specify the capacity of the LRU cache used by dnssec
middleware.
Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
Add auto-load middleware that automatically picks up zones.
Every X seconds it will scan for new zones.
Add tests and documentation.
Make 'make test' use -race.
In the file middleware, if there are multiple zone files under a single root zone, watchers are only invoked on the last element of `zones.Names`. This is caused by loop override on the variable `n`.
This issue can be fixed by passing zone object, which calls reload watcher directly.
* Replace go-cache with golang-lru
This fix replace go-cache with golang-lru, as is specified in 335.
Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
* Move cache initialization to setup
This commit move cache initialization to setup in dnssec middleware.
Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
This fix adds several needed comments in
`middleware/kubernetes/nametemplate/nametemplate.go`
to clean up golint output.
There are still 3 places that needs proper docs:
```
middleware/kubernetes/nametemplate/nametemplate.go:64:1: comment on exported type Template should be of the form "Template ..." (with optional leading article)
middleware/kubernetes/nametemplate/nametemplate.go:72:1: comment on exported method Template.SetTemplate should be of the form "SetTemplate ..."
middleware/kubernetes/nametemplate/nametemplate.go:188:1: comment on exported type NameValues should be of the form "NameValues ..." (with optional leading article)
```
Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
This PR adds the *root* middleware that specifies a path where
all zone file (the *file* middleware is the only consumer now) can
be found. It works the same as in Caddy.
Documentation can be found in the README.md of the middleware.
Fixes#307
This fix adds unit tests for `ExtractAddressFromReverse`. It also
updates the implementation of `ExtractAddressFromReverse` so that
only a valid IPv4 or IPv6 address (in string format) is returned.
If the addresses is not valid and can not be handled by golang's
`net.ParseIP` then an empty string (`""`) will be returned.
Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
This still needs cleanup, but this is a first pass the cleans some
cruft and documents our style (in middleware.md) and makes all the
docs match that style.
Rename: positive -> success
negative -> denial
There is a third (unused category) which is error. Start using these
new in the caching middleware and later in the logging middleware.
* middleware/proxy
Use connection pooling for communicating with an upstream, instead of
opening a new socket every time.
This makes the proxy more efficient and allowed for some cleanups.
* Some cleanups
* Some fixes
* more
* Kill pool
* Add nil check
* remove pool
Make DefaultPort just be 53. And use the value of Port to make the zone
stanzas complete (instead of defaulting to 53 always). This allows you
to override the port with dns.port.
Make the methods that handle Metrics all use pointer receivers to fix
sync.Once not being initialized.
Finish the setup_test to test for failures. And make the check for the
address more strict and return an error when it does not have a port
number.
Add a toplevel test that starts a CoreDNS server with metrics enabled
so we catch these errors in the future.
