Seena Fallah 736a2d58ae
feat!: generalize PTR record support from rfc2136 to all providers (#6232)
* feat(metrics): add source wrapper metrics for invalid and deduplicated endpoints

Add GaugeVecMetric.Reset() to clear stale label combinations between cycles.

Introduce invalidEndpoints and deduplicatedEndpoints gauge vectors in the
source wrappers package, partitioned by record_type and source_type. The
dedup source wrapper now tracks rejected (invalid) and de-duplicated
endpoints per collection cycle.

Update the metrics documentation and bump the known metrics count.

Signed-off-by: Seena Fallah <seenafallah@gmail.com>

* feat(source): add PTR source wrapper for automatic reverse DNS

Implement ptrSource, a source wrapper that generates PTR endpoints from
A/AAAA records. The wrapper supports:

- Global default via WithCreatePTR (maps to --create-ptr flag)
- Per-endpoint override via record-type provider-specific property
- Grouping multiple hostnames sharing an IP into a single PTR endpoint
- Skipping wildcard DNS names

Add WithPTRSupported and WithCreatePTR options to the wrapper Config
and wire the PTR wrapper into the WrapSources chain when PTR is in
managed-record-types.

Signed-off-by: Seena Fallah <seenafallah@gmail.com>

* feat(config): add --create-ptr flag and deprecate --rfc2136-create-ptr

Add the generic --create-ptr boolean flag to Config, enabling automatic
PTR record creation for any provider. Add IsPTRSupported() helper that
checks whether PTR is included in --managed-record-types.

Add validation: --create-ptr (or legacy --rfc2136-create-ptr) now
requires PTR in --managed-record-types, preventing misconfiguration.

Mark --rfc2136-create-ptr as deprecated in the flag description.

Signed-off-by: Seena Fallah <seenafallah@gmail.com>

* refactor(rfc2136): remove inline PTR logic in favor of PTR source wrapper

Remove the createPTR field, AddReverseRecord, RemoveReverseRecord, and
GenerateReverseRecord methods from the rfc2136 provider. PTR record
generation is now handled generically by the PTR source wrapper before
records reach the provider.

Update the PTR creation test to supply pre-generated PTR endpoints
(simulating what the source wrapper produces) instead of relying on
the provider to create them internally.

Signed-off-by: Seena Fallah <seenafallah@gmail.com>

* feat(controller): wire PTR source wrapper into buildSource

Pass the top-level Config to buildSource so it can read IsPTRSupported()
and the CreatePTR / RFC2136CreatePTR flags. When PTR is in
managed-record-types, the PTR source wrapper is installed in the
wrapper chain with the combined create-ptr default.

Signed-off-by: Seena Fallah <seenafallah@gmail.com>

* chore(pdns): remove stale comment and fix whitespace

Remove an outdated comment about a single-target-per-tuple assumption
that no longer applies.

Signed-off-by: Seena Fallah <seenafallah@gmail.com>

* docs: add PTR records documentation and update existing guides

Add docs/advanced/ptr-records.md covering the --create-ptr flag,
per-resource annotation overrides, prerequisites, and usage examples.

Update:
- annotations.md: document record-type annotation
- flags.md: add --create-ptr, mark --rfc2136-create-ptr as deprecated
- tutorials/rfc2136.md: point to generic --create-ptr flag
- contributing/source-wrappers.md: add PTR wrapper to the chain
- mkdocs.yml: add PTR Records navigation entry

Signed-off-by: Seena Fallah <seenafallah@gmail.com>

* feat(rfc2136)!: remove rfc2136-create-ptr in favor of create-ptr

Signed-off-by: Seena Fallah <seenafallah@gmail.com>

---------

Signed-off-by: Seena Fallah <seenafallah@gmail.com>
2026-03-30 13:36:16 +05:30

577 lines
18 KiB
Go

/*
Copyright 2018 The Kubernetes Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package pdns
import (
"bytes"
"context"
"crypto/tls"
"encoding/json"
"errors"
"math"
"net"
"net/http"
"slices"
"sort"
"strings"
"time"
pgo "github.com/ffledgling/pdns-go"
log "github.com/sirupsen/logrus"
"sigs.k8s.io/external-dns/pkg/apis/externaldns"
"sigs.k8s.io/external-dns/endpoint"
"sigs.k8s.io/external-dns/pkg/tlsutils"
"sigs.k8s.io/external-dns/plan"
"sigs.k8s.io/external-dns/provider"
)
type pdnsChangeType string
const (
apiBase = "/api/v1"
defaultTTL = 300
// PdnsDelete and PdnsReplace are effectively an enum for "pgo.RrSet.changetype"
// TODO: Can we somehow get this from the pgo swagger client library itself?
// PdnsDelete : PowerDNS changetype used for deleting rrsets
// ref: https://doc.powerdns.com/authoritative/http-api/zone.html#rrset (see "changetype")
PdnsDelete pdnsChangeType = "DELETE"
// PdnsReplace : PowerDNS changetype for creating, updating and patching rrsets
PdnsReplace pdnsChangeType = "REPLACE"
// Number of times to retry failed PDNS requests
retryLimit = 3
// time in milliseconds
retryAfterTime = 250 * time.Millisecond
)
// record types which require to have trailing dot
var trailingTypes = []string{
endpoint.RecordTypeCNAME,
endpoint.RecordTypeMX,
endpoint.RecordTypeSRV,
endpoint.RecordTypeNS,
endpoint.RecordTypePTR,
"ALIAS",
}
// PDNSConfig is comprised of the fields necessary to create a new PDNSProvider
type PDNSConfig struct {
DomainFilter *endpoint.DomainFilter
DryRun bool
Server string
ServerID string
APIKey string
TLSConfig TLSConfig
}
// TLSConfig is comprised of the TLS-related fields necessary to create a new PDNSProvider
type TLSConfig struct {
SkipTLSVerify bool
CAFilePath string
ClientCertFilePath string
ClientCertKeyFilePath string
}
func (tlsConfig *TLSConfig) setHTTPClient(pdnsClientConfig *pgo.Configuration) error {
log.Debug("Configuring TLS for PDNS Provider.")
tlsClientConfig, err := tlsutils.NewTLSConfig(
tlsConfig.ClientCertFilePath,
tlsConfig.ClientCertKeyFilePath,
tlsConfig.CAFilePath,
"",
tlsConfig.SkipTLSVerify,
tls.VersionTLS12,
)
if err != nil {
return err
}
// Timeouts taken from net.http.DefaultTransport
transporter := &http.Transport{
Proxy: http.ProxyFromEnvironment,
DialContext: (&net.Dialer{
Timeout: 30 * time.Second,
KeepAlive: 30 * time.Second,
DualStack: true,
}).DialContext,
MaxIdleConns: 100,
IdleConnTimeout: 90 * time.Second,
TLSHandshakeTimeout: 10 * time.Second,
ExpectContinueTimeout: 1 * time.Second,
TLSClientConfig: tlsClientConfig,
}
pdnsClientConfig.HTTPClient = &http.Client{
Transport: transporter,
}
return nil
}
// Function for debug printing
func stringifyHTTPResponseBody(r *http.Response) string {
if r == nil {
return ""
}
buf := new(bytes.Buffer)
_, _ = buf.ReadFrom(r.Body)
return buf.String()
}
// PDNSAPIProvider : Interface used and extended by the PDNSAPIClient struct as
// well as mock APIClients used in testing
type PDNSAPIProvider interface {
ListZones() ([]pgo.Zone, *http.Response, error)
ListZone(zoneID string) (pgo.Zone, *http.Response, error)
PatchZone(zoneID string, zoneStruct pgo.Zone) (*http.Response, error)
}
// PDNSAPIClient : Struct that encapsulates all the PowerDNS specific implementation details
type PDNSAPIClient struct {
dryRun bool
serverID string
authCtx context.Context
client *pgo.APIClient
}
// ListZones : Method returns all enabled zones from PowerDNS
// ref: https://doc.powerdns.com/authoritative/http-api/zone.html#get--servers-server_id-zones
func (c *PDNSAPIClient) ListZones() ([]pgo.Zone, *http.Response, error) {
var zones []pgo.Zone
var resp *http.Response
var err error
for i := range retryLimit {
zones, resp, err = c.client.ZonesApi.ListZones(c.authCtx, c.serverID)
if err != nil {
log.Debugf("Unable to fetch zones %v", err)
log.Debugf("Retrying ListZones() ... %d", i)
time.Sleep(retryAfterTime * (1 << uint(i)))
continue
}
return zones, resp, err
}
return zones, resp, provider.NewSoftErrorf("unable to list zones: %v", err)
}
// partitionZones returns a slice of zones that adhere to the domain filter and a slice of ones that do not adhere to the filter.
func partitionZones(zones []pgo.Zone, domainFilter *endpoint.DomainFilter) ([]pgo.Zone, []pgo.Zone) {
if domainFilter == nil || !domainFilter.IsConfigured() {
return zones, nil
}
var filtered, residual []pgo.Zone
for _, zone := range zones {
if domainFilter.Match(zone.Name) {
filtered = append(filtered, zone)
} else {
residual = append(residual, zone)
}
}
return filtered, residual
}
// ListZone : Method returns the details of a specific zone from PowerDNS
// ref: https://doc.powerdns.com/authoritative/http-api/zone.html#get--servers-server_id-zones-zone_id
func (c *PDNSAPIClient) ListZone(zoneID string) (pgo.Zone, *http.Response, error) {
for i := range retryLimit {
zone, resp, err := c.client.ZonesApi.ListZone(c.authCtx, c.serverID, zoneID)
if err != nil {
log.Debugf("Unable to fetch zone %v", err)
log.Debugf("Retrying ListZone() ... %d", i)
time.Sleep(retryAfterTime * (1 << uint(i)))
continue
}
return zone, resp, err
}
return pgo.Zone{}, nil, provider.NewSoftErrorf("unable to list zone")
}
// PatchZone : Method used to update the contents of a particular zone from PowerDNS
// ref: https://doc.powerdns.com/authoritative/http-api/zone.html#patch--servers-server_id-zones-zone_id
func (c *PDNSAPIClient) PatchZone(zoneID string, zoneStruct pgo.Zone) (*http.Response, error) {
var resp *http.Response
var err error
for i := range retryLimit {
resp, err = c.client.ZonesApi.PatchZone(c.authCtx, c.serverID, zoneID, zoneStruct)
if err != nil {
log.Debugf("Unable to patch zone %v", err)
log.Debugf("Retrying PatchZone() ... %d", i)
time.Sleep(retryAfterTime * (1 << uint(i)))
continue
}
return resp, err
}
return resp, provider.NewSoftErrorf("unable to patch zone: %v", err)
}
// PDNSProvider is an implementation of the Provider interface for PowerDNS
type PDNSProvider struct {
provider.BaseProvider
client PDNSAPIProvider
domainFilter *endpoint.DomainFilter
}
// New creates a PowerDNS provider from the given configuration.
func New(ctx context.Context, cfg *externaldns.Config, domainFilter *endpoint.DomainFilter) (provider.Provider, error) {
return newProvider(
ctx,
PDNSConfig{
DomainFilter: domainFilter,
DryRun: cfg.DryRun,
Server: cfg.PDNSServer,
ServerID: cfg.PDNSServerID,
APIKey: cfg.PDNSAPIKey,
TLSConfig: TLSConfig{
SkipTLSVerify: cfg.PDNSSkipTLSVerify,
CAFilePath: cfg.TLSCA,
ClientCertFilePath: cfg.TLSClientCert,
ClientCertKeyFilePath: cfg.TLSClientCertKey,
},
},
)
}
// newProvider initializes a new PowerDNS based Provider.
func newProvider(ctx context.Context, config PDNSConfig) (*PDNSProvider, error) {
// Do some input validation
if config.APIKey == "" {
return nil, errors.New("missing API Key for PDNS. Specify using --pdns-api-key=")
}
// We do not support dry running, exit safely instead of surprising the user
// TODO: Add Dry Run support
if config.DryRun {
return nil, errors.New("PDNS Provider does not currently support dry-run")
}
if config.Server == "localhost" {
log.Warnf("PDNS Server is set to localhost, this may not be what you want. Specify using --pdns-server=")
}
pdnsClientConfig := pgo.NewConfiguration()
pdnsClientConfig.BasePath = config.Server + apiBase
if err := config.TLSConfig.setHTTPClient(pdnsClientConfig); err != nil {
return nil, err
}
provider := &PDNSProvider{
client: &PDNSAPIClient{
dryRun: config.DryRun,
serverID: config.ServerID,
authCtx: context.WithValue(ctx, pgo.ContextAPIKey, pgo.APIKey{Key: config.APIKey}),
client: pgo.NewAPIClient(pdnsClientConfig),
},
domainFilter: config.DomainFilter,
}
return provider, nil
}
// filteredZones fetches all zones from the PowerDNS API and partitions them
// using the provider's domain filter. It returns the matching zones, the
// non-matching (residual) zones, and any error from the API call.
func (p *PDNSProvider) filteredZones() ([]pgo.Zone, []pgo.Zone, error) {
zones, _, err := p.client.ListZones()
if err != nil {
return nil, nil, err
}
filtered, residual := partitionZones(zones, p.domainFilter)
return filtered, residual, nil
}
func (p *PDNSProvider) GetDomainFilter() endpoint.DomainFilterInterface {
// Return all zones the provider manages so the controller can intersect
// with --domain-filter on its own. Do NOT apply p.domainFilter here;
// double-filtering would produce an empty filter when no zones match,
// silently failing open instead of letting the controller see the
// mismatch and produce a safe empty plan.
zones, _, err := p.client.ListZones()
if err != nil {
log.Errorf("Unable to fetch zones from PowerDNS API: %v", err)
return &endpoint.DomainFilter{}
}
zoneNames := make([]string, 0, 2*len(zones))
for _, zone := range zones {
zoneNames = append(zoneNames, zone.Name, "."+zone.Name)
}
return endpoint.NewDomainFilter(zoneNames)
}
// hasAliasAnnotation checks if the endpoint has the alias annotation set to true
func (p *PDNSProvider) hasAliasAnnotation(ep *endpoint.Endpoint) bool {
value, exists := ep.GetProviderSpecificProperty("alias")
return exists && value == "true"
}
func (p *PDNSProvider) convertRRSetToEndpoints(rr pgo.RrSet) []*endpoint.Endpoint {
endpoints := make([]*endpoint.Endpoint, 0)
targets := make([]string, 0)
rrType_ := rr.Type_
for _, record := range rr.Records {
// If a record is "Disabled", it's not supposed to be "visible"
if !record.Disabled {
targets = append(targets, record.Content)
}
}
if rr.Type_ == "ALIAS" {
rrType_ = endpoint.RecordTypeCNAME
}
endpoints = append(endpoints, endpoint.NewEndpointWithTTL(rr.Name, rrType_, endpoint.TTL(rr.Ttl), targets...))
return endpoints
}
// ConvertEndpointsToZones marshals endpoints into pdns compatible Zone structs
func (p *PDNSProvider) ConvertEndpointsToZones(eps []*endpoint.Endpoint, changetype pdnsChangeType) ([]pgo.Zone, error) {
var zoneList = make([]pgo.Zone, 0)
endpoints := make([]*endpoint.Endpoint, len(eps))
copy(endpoints, eps)
// Sort the endpoints array so we have deterministic inserts
sort.SliceStable(endpoints,
func(i, j int) bool {
// We only care about sorting endpoints with the same dnsname
if endpoints[i].DNSName == endpoints[j].DNSName {
return endpoints[i].RecordType < endpoints[j].RecordType
}
return endpoints[i].DNSName < endpoints[j].DNSName
})
filteredZones, residualZones, err := p.filteredZones()
if err != nil {
return nil, err
}
// Sort the zone by length of the name in descending order, we use this
// property later to ensure we add a record to the longest matching zone
sort.SliceStable(filteredZones, func(i, j int) bool { return len(filteredZones[i].Name) > len(filteredZones[j].Name) })
// NOTE: Complexity of this loop is O(FilteredZones*Endpoints).
// A possibly faster implementation would be a search of the reversed
// DNSName in a trie of Zone names, which should be O(Endpoints), but at this point it's not
// necessary.
for _, zone := range filteredZones {
zone.Rrsets = []pgo.RrSet{}
for i := 0; i < len(endpoints); {
ep := endpoints[i]
dnsname := provider.EnsureTrailingDot(ep.DNSName)
if dnsname == zone.Name || strings.HasSuffix(dnsname, "."+zone.Name) {
records := []pgo.Record{}
RecordType_ := ep.RecordType
for _, t := range ep.Targets {
if slices.Contains(trailingTypes, ep.RecordType) {
t = provider.EnsureTrailingDot(t)
}
records = append(records, pgo.Record{Content: t})
}
// Check if we should use ALIAS instead of CNAME:
// 1. APEX records (dnsname == zone.Name) always use ALIAS
// 2. If annotation external-dns.alpha.kubernetes.io/alias=true is set
// (can be set via --prefer-alias flag globally or per-resource annotation)
if ep.RecordType == endpoint.RecordTypeCNAME {
useAlias := dnsname == zone.Name || p.hasAliasAnnotation(ep)
if useAlias {
log.Debugf("Converting CNAME record %q to ALIAS", dnsname)
RecordType_ = "ALIAS"
}
}
rrset := pgo.RrSet{
Name: dnsname,
Type_: RecordType_,
Records: records,
Changetype: string(changetype),
}
// DELETEs explicitly forbid a TTL, therefore only PATCHes need the TTL
if changetype == PdnsReplace {
if int64(ep.RecordTTL) > int64(math.MaxInt32) {
return nil, provider.NewSoftErrorf("value of record TTL overflows, limited to int32")
}
if ep.RecordTTL == 0 {
// No TTL was specified for the record, we use the default
rrset.Ttl = int32(defaultTTL)
} else {
rrset.Ttl = int32(ep.RecordTTL)
}
}
zone.Rrsets = append(zone.Rrsets, rrset)
// "pop" endpoint if it's matched
endpoints = append(endpoints[0:i], endpoints[i+1:]...)
} else {
// If we didn't pop anything, we move to the next item in the list
i++
}
}
if len(zone.Rrsets) > 0 {
zoneList = append(zoneList, zone)
}
}
// residualZones is unsorted by name length like its counterpart
// since we only care to remove endpoints that do not match domain filter
for _, zone := range residualZones {
for i := 0; i < len(endpoints); {
ep := endpoints[i]
dnsname := provider.EnsureTrailingDot(ep.DNSName)
if dnsname == zone.Name || strings.HasSuffix(dnsname, "."+zone.Name) {
// "pop" endpoint if it's matched to a residual zone... essentially a no-op
log.Debugf("Ignoring Endpoint because it was matched to a zone that was not specified within Domain Filter(s): %s", dnsname)
endpoints = append(endpoints[0:i], endpoints[i+1:]...)
} else {
i++
}
}
}
// If we still have some endpoints left, it means we couldn't find a matching zone (filtered or residual) for them
// We warn instead of hard fail here because we don't want a misconfig to cause everything to go down
if len(endpoints) > 0 {
log.Warnf("No matching zones were found for the following endpoints: %+v", endpoints)
}
log.Debugf("Zone List generated from Endpoints: %+v", zoneList)
return zoneList, nil
}
// mutateRecords takes a list of endpoints and creates, replaces or deletes them based on the changetype
func (p *PDNSProvider) mutateRecords(endpoints []*endpoint.Endpoint, changetype pdnsChangeType) error {
zonelist, err := p.ConvertEndpointsToZones(endpoints, changetype)
if err != nil {
return err
}
for _, zone := range zonelist {
jso, err := json.Marshal(zone)
if err != nil {
log.Errorf("JSON Marshal for zone struct failed!")
} else {
log.Debugf("Struct for PatchZone:\n%s", string(jso))
}
resp, err := p.client.PatchZone(zone.Id, zone)
if err != nil {
log.Debugf("PDNS API response: %s", stringifyHTTPResponseBody(resp))
return err
}
}
return nil
}
// Records returns all DNS records controlled by the configured PDNS server (for all zones)
func (p *PDNSProvider) Records(_ context.Context) ([]*endpoint.Endpoint, error) {
filteredZones, _, err := p.filteredZones()
if err != nil {
return nil, err
}
var endpoints []*endpoint.Endpoint
for _, zone := range filteredZones {
z, _, err := p.client.ListZone(zone.Id)
if err != nil {
return nil, provider.NewSoftErrorf("unable to fetch records: %v", err)
}
for _, rr := range z.Rrsets {
endpoints = append(endpoints, p.convertRRSetToEndpoints(rr)...)
}
}
log.Debugf("Records fetched:\n%+v", endpoints)
return endpoints, nil
}
// AdjustEndpoints performs checks on the provided endpoints and will skip any potentially failing changes.
func (p *PDNSProvider) AdjustEndpoints(endpoints []*endpoint.Endpoint) ([]*endpoint.Endpoint, error) {
var validEndpoints []*endpoint.Endpoint
for i := range endpoints {
if !endpoints[i].CheckEndpoint() {
log.Warnf("Ignoring Endpoint because of invalid %v record formatting: {Target: '%v'}", endpoints[i].RecordType, endpoints[i].Targets)
continue
}
validEndpoints = append(validEndpoints, endpoints[i])
}
return validEndpoints, nil
}
// ApplyChanges takes a list of changes (endpoints) and updates the PDNS server
// by sending the correct HTTP PATCH requests to a matching zone
func (p *PDNSProvider) ApplyChanges(_ context.Context, changes *plan.Changes) error {
startTime := time.Now()
// Create
for _, change := range changes.Create {
log.Infof("CREATE: %+v", change)
}
// We only attempt to mutate records if there are any to mutate. A
// call to mutate records with an empty list of endpoints is still a
// valid call and a no-op, but we might as well not make the call to
// prevent unnecessary logging
if len(changes.Create) > 0 {
// "Replacing" non-existent records creates them
err := p.mutateRecords(changes.Create, PdnsReplace)
if err != nil {
return err
}
}
// Update
for _, change := range changes.UpdateOld {
// Since PDNS "Patches", we don't need to specify the "old"
// record. The Update New change type will automatically take
// care of replacing the old RRSet with the new one We simply
// leave this logging here for information
log.Debugf("UPDATE-OLD (ignored): %+v", change)
}
for _, change := range changes.UpdateNew {
log.Infof("UPDATE-NEW: %+v", change)
}
if len(changes.UpdateNew) > 0 {
err := p.mutateRecords(changes.UpdateNew, PdnsReplace)
if err != nil {
return err
}
}
// Delete
for _, change := range changes.Delete {
log.Infof("DELETE: %+v", change)
}
if len(changes.Delete) > 0 {
err := p.mutateRecords(changes.Delete, PdnsDelete)
if err != nil {
return err
}
}
log.Infof("Changes pushed out to PowerDNS in %s\n", time.Since(startTime))
return nil
}