These are too hidden now. They increase the issue-load, because people
don't see them.
Add log.Errorf in the core/dnsserver recover routine.
Signed-off-by: Miek Gieben <miek@miek.nl>
Move github.com/openzipkin/zipkin-go-opentracing => github.com/openzipkin-contrib/zipkin-go-opentracing
and run `go mod tidy`
Fixes 3069
Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
I don't believe this is actually needed (anymore). The:
os.Stderr = os.Stdout
is a crazy hack that def. needs to go.
Signed-off-by: Miek Gieben <miek@miek.nl>
* contrib: add words on adding new plugins
New plugins are usually dropped in one giant ball of code, add some
words on how this can be approached and made to work faster.
(will at least give us something to point at)
Signed-off-by: Miek Gieben <miek@miek.nl>
* Update CONTRIBUTING.md
Co-Authored-By: Michael Grosser <development@stp-ip.net>
* Update CONTRIBUTING.md
Co-Authored-By: Michael Grosser <development@stp-ip.net>
* Update CONTRIBUTING.md
Co-Authored-By: Michael Grosser <development@stp-ip.net>
* Update CONTRIBUTING.md
* Update CONTRIBUTING.md
Co-Authored-By: Michael Grosser <development@stp-ip.net>
* Fix corefile usage
* plugin/file: allow README.md testing
Allow readme testing for the file plugin and fix bugs that where found:
* the reader wasn't reset when re-reading the same io.reader for a
different origin.
Signed-off-by: Miek Gieben <miek@miek.nl>
* Update test/example_test.go
Co-Authored-By: Michael Grosser <development@stp-ip.net>
The host plugin kept on adding entries instead of overwriting. Split the
inline cache off from the /etc/hosts file cache and clear /etc/hosts
file cache and re-parsing.
A bunch of other cleanup as well. Use functions defined in the plugin
package, don't re-parse strings if you don't have to and use To4() to
check the family for IP addresses. Fix all test cases a carried entries
are always fqdn-ed. Various smaller cleanup in unnessacry constants.
Fixes: #3014
Signed-off-by: Miek Gieben <miek@miek.nl>
* plugin/file: simplify locking
Simplify the locking, remove the reloadMu and just piggyback on the
other lock for accessing content, which assumes things can be move
underneath.
Copy the Apex and Zone to new vars to make sure the pointer isn't
updated from under us.
The releadMu isn't need at all, the time.Ticker firing while we're
reading means we will just miss that tick and get it on the next go.
Add rrutil subpackage and put some more generic functions in there, that
are now used from file and the tree package. This removes some
duplication.
Rename additionalProcessing that didn't actually do that to
externalLookup, because that's what being done at some point.
Signed-off-by: Miek Gieben <miek@miek.nl>
* Update plugin/file/lookup.go
Co-Authored-By: Michael Grosser <development@stp-ip.net>
After calling NewZone the number of records should be zero, but due to
how zone.All() was implemented so empty RRs would be added. This then
fails the == 0 check in xfr.go and put nil in the slice, this then
subsequently panics on the Len().
Fix this making All() smarter when adding records. Added little test to
enfore this.
Signed-off-by: Miek Gieben <miek@miek.nl>
* pkg/cache: fix race in Add() and Evict()
This fixes a race in Add() when the shard is at max capacity and the key
being added is already stored. Previously, the shard would evict a
random value - when all it needed to do was replace an existing value.
There was a race in how Evict() picked which key to remove, which would
cause concurrent calls to Evict() to remove the same key.
Additionally, this commit removes a lot of the lock contention and a
race around Add() and Evict() by changing them to immediately hold the
write lock. Previously, they would check conditions with the read lock
held and not re-check those conditions once the write lock was acquired
(this is a race).
* pkg/cache: code review comments
* pkg/cache: simplify Add() logic
* plugin/file: Rename do to walk, cleanup and document
* This renames Do to Walk to be more inline with Go standards. Also make
it return an error instead of a bool.
Also give give walk access to rrs. Alternatively e.m could be
exported, but just access the map of rrs should work as well. Another
alternative would be adding a whole bunch of helper functions, but
those need grab and return the data. Just having access to the rrs
should be easiest for most Walks.
* It adds Type and TypeForWildcard to show the different functions
* *Removes* the identical RR check when inserting; this was only done
for A, AAAA and MX and not finished; removed under the mantra garbage
in garbage out.
* Reuses Types to return all the types in an *tree.Elem
Signed-off-by: Miek Gieben <miek@miek.nl>
* better comments
Signed-off-by: Miek Gieben <miek@miek.nl>
* plugin/file: fix setting ReloadInterval
The reload interval was only correctly set if there was an extra
block for the file. Move this down to set up.
Add test case that fails before, but now works.
Signed-off-by: Miek Gieben <miek@miek.nl>
* layout and use Errorf
Signed-off-by: Miek Gieben <miek@miek.nl>
* replacer: evaluate format once and improve perf by ~3x
This improves the performance of logging by almost 3x and reduces memory
usage by ~8x.
Benchmark results:
benchmark old ns/op new ns/op delta
BenchmarkReplacer-12 644 324 -49.69%
BenchmarkReplacer_CommonLogFormat-12 4228 1471 -65.21%
benchmark old allocs new allocs delta
BenchmarkReplacer-12 8 2 -75.00%
BenchmarkReplacer_CommonLogFormat-12 51 17 -66.67%
benchmark old bytes new bytes delta
BenchmarkReplacer-12 240 48 -80.00%
BenchmarkReplacer_CommonLogFormat-12 3723 446 -88.02%
* replacer: code review comments
* bufPool: document why we use a pointer to a slice
* parseFormat: fix confusing comment
* TestParseFormat_Nodes: rename to TestParseFormatNodes
* replacer: use a value for bufPool instead of a pointer
* replacer: remove comment
* replacer: replace labels with knownLabels
The previous slice of label names is no longer needed.
* log: use ioutil.Discard as write buffer in benchmark
Using a buffer gives unrealistic stats and consumes a large amount of
memory.
* log: lazily check if a msg should be logged
* log: improve variable name
Change 'ok' to the more descriptive 'shouldLog'.
* log: code comments: don't reuse variable
miekg/dns recently became more aggressive in setting TC; anything that's
chopped of a response results in TC. Amend the tests. Disputable if
these TC checks still add something.
Signed-off-by: Miek Gieben <miek@miek.nl>
Ditch our truncation code and use the upstream one in miekg/dns.
This saves code on our end, end upstream is also more efficient as every
RR is Len-ed only once. With our bin-search this is not guaranteed.
Signed-off-by: Miek Gieben <miek@miek.nl>
This fix fixes the `\ No newline at end of file`
in plugin/chaos/zowners.go, by adding `"\n"`
to the end of owners_generate.go.
Also fixes a gofmt issue in plugin/etcd/setup.go
Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
This fix updates:
- github.com/miekg/dns to 1.15 (#2947)
- github.com/aws/aws-sdk-go to 1.20.13 (#2953)
- google.golang.org/grpc to 1.22.0 (#2966)
This fix fixes#2947
This fix fixes#2953
This fix fixes#2966
Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
* plugin/file: make non-existent file non-fatal
If the zone file being loaded doesn't exist *and* reload is enabled,
just wait the file to pop up in the normal Reload routine.
If reload is set to 0s; we keep this a fatal error on startup. Aslo fix
the ticker in z.Reload(): remove the per second ticks and just use the
reload interval for the ticker.
Brush up the documentation a bit as well.
Fixes: #2951
Signed-off-by: Miek Gieben <miek@miek.nl>
* Stickler and test compile
Signed-off-by: Miek Gieben <miek@miek.nl>
* Remove there too
Signed-off-by: Miek Gieben <miek@miek.nl>
* Cant README test these because zone files dont exist
Signed-off-by: Miek Gieben <miek@miek.nl>
