deal with not responding nameservers (offline or unreachable)

if a nameserver does not respond, we flag it in the database
(available = False) and stop talking to it for 5 minutes.

after 5 minutes, we flag it available = True again and retry talking to it
(and if not successful, flag it as unavailable again).

this avoids a unresponsive UI as we do a lot of ns queries from the templates
to get v4 and v6 addresses of our hosts from master nameservers.

we log a warning when setting unavailable and a info when setting available.

nsupdate/main/dnstools.py

@@ -4,6 +4,18 @@ Misc. DNS related code: query, dynamic update, etc.
 Usually, higher level code wants to call the add/update/delete functions.
 """
 
+# time to wait for dns name resolving [s]
+RESOLVER_TIMEOUT = 5.0
+
+# time to wait for dns name updating [s]
+UPDATE_TIMEOUT = 20.0
+
+# time after we retry to reach a previously unreachable ns [s]
+UNAVAILABLE_RETRY = 300.0
+
+
+from datetime import timedelta
+
 import logging
 logger = logging.getLogger(__name__)
 
@@ -16,6 +28,7 @@ import dns.tsig
 import dns.tsigkeyring
 
 from django.conf import settings
+from django.utils.timezone import now
 
 
 class SameIpError(ValueError):
@@ -25,6 +38,13 @@ class SameIpError(ValueError):
     """
 
 
+class NameServerNotAvailable(Exception):
+    """
+    raised if some nameserver was flagged as not available,
+    but we tried using it.
+    """
+
+
 def check_ip(ipaddr, keys=('ipv4', 'ipv6')):
     """
     Check if a string is a valid ip address and also
@@ -128,7 +148,7 @@ def query_ns(qname, rdtype, origin=None):
     :type rdtype: int or str
     :param origin: origin zone
     :type origin: str or None
-    :return: IP (as str)
+    :return: IP (as str) or "-" if ns is not available
     """
     origin, name = parse_name(qname, origin)
     origin_str = str(origin)
@@ -138,8 +158,16 @@ def query_ns(qname, rdtype, origin=None):
     # want into the documented attributes:
     resolver.nameservers = [nameserver, ]
     resolver.search = [dns.name.from_text(settings.BASEDOMAIN), ]
-    answer = resolver.query(qname, rdtype)
+    resolver.lifetime = RESOLVER_TIMEOUT
-    return str(list(answer)[0])
+    try:
+        answer = resolver.query(qname, rdtype)
+        ip = str(list(answer)[0])
+        return ip
+    except (dns.resolver.Timeout, dns.resolver.NoNameservers):  # socket.error also?
+        logger.warning("timeout when querying for name '%s' in zone '%s' with rdtype '%s'." % (
+                       name, origin, rdtype))
+        set_ns_availability(origin, False)
+        raise
 
 
 def parse_name(fqdn, origin=None):
@@ -169,9 +197,20 @@ def get_ns_info(origin):
 
     :param origin: zone we are dealing with, must be with trailing dot
     :return: master nameserver, update key, update algo
+    :raises: NameServerNotAvailable if ns was flagged unavailable in the db
     """
     from .models import Domain
-    d = Domain.objects.get(domain=origin.rstrip('.'))
+    domain = origin.rstrip('.')
+    d = Domain.objects.get(domain=domain)
+    if not d.available:
+        if d.last_update + timedelta(seconds=UNAVAILABLE_RETRY) > now():
+            # if there are troubles with a nameserver, we set available=False
+            # and stop trying working with that nameserver for a while
+            raise NameServerNotAvailable("nameserver for domain %s at IP %s was flagged unavailable" % (
+                                         domain, d.nameserver_ip, ))
+        else:
+            # retry timeout is over, set it available again
+            set_ns_availability(origin, True)
     algorithm = getattr(dns.tsig, d.nameserver_update_algorithm)
     return d.nameserver_ip, d.nameserver_update_key, algorithm
 
@@ -205,5 +244,31 @@ def update_ns(fqdn, rdtype='A', ipaddr=None, origin=None, action='upd', ttl=60):
         upd.replace(name, ttl, rdtype, ipaddr)
     logger.debug("performing %s for name %s and origin %s with rdtype %s and ipaddr %s" % (
                  action, name, origin, rdtype, ipaddr))
-    response = dns.query.tcp(upd, nameserver)
+    try:
-    return response
+        response = dns.query.tcp(upd, nameserver, timeout=UPDATE_TIMEOUT)
+        return response
+    except dns.exception.Timeout:
+        logger.warning("timeout when performing %s for name %s and origin %s with rdtype %s and ipaddr %s" % (
+                       action, name, origin, rdtype, ipaddr))
+        set_ns_availability(origin, False)
+        raise
+
+
+def set_ns_availability(domain, available):
+    """
+    Set availability of the master nameserver for <domain>.
+
+    As each Timeout takes quite a while, we want to avoid it.
+
+    :param domain: domain object or string, may end with "."
+    :param available: True/False for availability of ns
+    """
+    from .models import Domain
+    domain = str(domain).rstrip('.')
+    d = Domain.objects.get(domain=domain)
+    d.available = available
+    d.save()
+    if available:
+        logger.info("set zone '%s' to available" % domain)
+    else:
+        logger.warning("set zone '%s' to unavailable" % domain)

nsupdate/main/migrations/0012_auto__add_field_domain_available.py

@@ -0,0 +1,94 @@
+# -*- coding: utf-8 -*-
+import datetime
+from south.db import db
+from south.v2 import SchemaMigration
+from django.db import models
+
+
+class Migration(SchemaMigration):
+
+    def forwards(self, orm):
+        # Adding field 'Domain.available'
+        db.add_column(u'main_domain', 'available',
+                      self.gf('django.db.models.fields.BooleanField')(default=True),
+                      keep_default=False)
+
+
+    def backwards(self, orm):
+        # Deleting field 'Domain.available'
+        db.delete_column(u'main_domain', 'available')
+
+
+    models = {
+        u'auth.group': {
+            'Meta': {'object_name': 'Group'},
+            u'id': ('django.db.models.fields.AutoField', [], {'primary_key': 'True'}),
+            'name': ('django.db.models.fields.CharField', [], {'unique': 'True', 'max_length': '80'}),
+            'permissions': ('django.db.models.fields.related.ManyToManyField', [], {'to': u"orm['auth.Permission']", 'symmetrical': 'False', 'blank': 'True'})
+        },
+        u'auth.permission': {
+            'Meta': {'ordering': "(u'content_type__app_label', u'content_type__model', u'codename')", 'unique_together': "((u'content_type', u'codename'),)", 'object_name': 'Permission'},
+            'codename': ('django.db.models.fields.CharField', [], {'max_length': '100'}),
+            'content_type': ('django.db.models.fields.related.ForeignKey', [], {'to': u"orm['contenttypes.ContentType']"}),
+            u'id': ('django.db.models.fields.AutoField', [], {'primary_key': 'True'}),
+            'name': ('django.db.models.fields.CharField', [], {'max_length': '50'})
+        },
+        u'auth.user': {
+            'Meta': {'object_name': 'User'},
+            'date_joined': ('django.db.models.fields.DateTimeField', [], {'default': 'datetime.datetime.now'}),
+            'email': ('django.db.models.fields.EmailField', [], {'max_length': '75', 'blank': 'True'}),
+            'first_name': ('django.db.models.fields.CharField', [], {'max_length': '30', 'blank': 'True'}),
+            'groups': ('django.db.models.fields.related.ManyToManyField', [], {'to': u"orm['auth.Group']", 'symmetrical': 'False', 'blank': 'True'}),
+            u'id': ('django.db.models.fields.AutoField', [], {'primary_key': 'True'}),
+            'is_active': ('django.db.models.fields.BooleanField', [], {'default': 'True'}),
+            'is_staff': ('django.db.models.fields.BooleanField', [], {'default': 'False'}),
+            'is_superuser': ('django.db.models.fields.BooleanField', [], {'default': 'False'}),
+            'last_login': ('django.db.models.fields.DateTimeField', [], {'default': 'datetime.datetime.now'}),
+            'last_name': ('django.db.models.fields.CharField', [], {'max_length': '30', 'blank': 'True'}),
+            'password': ('django.db.models.fields.CharField', [], {'max_length': '128'}),
+            'user_permissions': ('django.db.models.fields.related.ManyToManyField', [], {'to': u"orm['auth.Permission']", 'symmetrical': 'False', 'blank': 'True'}),
+            'username': ('django.db.models.fields.CharField', [], {'unique': 'True', 'max_length': '30'})
+        },
+        u'contenttypes.contenttype': {
+            'Meta': {'ordering': "('name',)", 'unique_together': "(('app_label', 'model'),)", 'object_name': 'ContentType', 'db_table': "'django_content_type'"},
+            'app_label': ('django.db.models.fields.CharField', [], {'max_length': '100'}),
+            u'id': ('django.db.models.fields.AutoField', [], {'primary_key': 'True'}),
+            'model': ('django.db.models.fields.CharField', [], {'max_length': '100'}),
+            'name': ('django.db.models.fields.CharField', [], {'max_length': '100'})
+        },
+        u'main.blacklisteddomain': {
+            'Meta': {'object_name': 'BlacklistedDomain'},
+            'created': ('django.db.models.fields.DateTimeField', [], {'auto_now_add': 'True', 'blank': 'True'}),
+            'created_by': ('django.db.models.fields.related.ForeignKey', [], {'to': u"orm['auth.User']", 'null': 'True', 'blank': 'True'}),
+            'domain': ('django.db.models.fields.CharField', [], {'unique': 'True', 'max_length': '256'}),
+            u'id': ('django.db.models.fields.AutoField', [], {'primary_key': 'True'}),
+            'last_update': ('django.db.models.fields.DateTimeField', [], {'auto_now': 'True', 'blank': 'True'})
+        },
+        u'main.domain': {
+            'Meta': {'object_name': 'Domain'},
+            'available': ('django.db.models.fields.BooleanField', [], {'default': 'True'}),
+            'available_for_everyone': ('django.db.models.fields.BooleanField', [], {'default': 'False'}),
+            'created': ('django.db.models.fields.DateTimeField', [], {'auto_now_add': 'True', 'blank': 'True'}),
+            'created_by': ('django.db.models.fields.related.ForeignKey', [], {'to': u"orm['auth.User']", 'null': 'True', 'blank': 'True'}),
+            'domain': ('django.db.models.fields.CharField', [], {'unique': 'True', 'max_length': '256'}),
+            u'id': ('django.db.models.fields.AutoField', [], {'primary_key': 'True'}),
+            'last_update': ('django.db.models.fields.DateTimeField', [], {'auto_now': 'True', 'blank': 'True'}),
+            'nameserver_ip': ('django.db.models.fields.GenericIPAddressField', [], {'max_length': '39'}),
+            'nameserver_update_algorithm': ('django.db.models.fields.CharField', [], {'default': "'HMAC_SHA512'", 'max_length': '256'}),
+            'nameserver_update_key': ('django.db.models.fields.CharField', [], {'max_length': '256'})
+        },
+        u'main.host': {
+            'Meta': {'unique_together': "(('subdomain', 'domain'),)", 'object_name': 'Host'},
+            'comment': ('django.db.models.fields.CharField', [], {'default': "''", 'max_length': '256', 'null': 'True', 'blank': 'True'}),
+            'created': ('django.db.models.fields.DateTimeField', [], {'auto_now_add': 'True', 'blank': 'True'}),
+            'created_by': ('django.db.models.fields.related.ForeignKey', [], {'related_name': "'hosts'", 'to': u"orm['auth.User']"}),
+            'domain': ('django.db.models.fields.related.ForeignKey', [], {'to': u"orm['main.Domain']"}),
+            u'id': ('django.db.models.fields.AutoField', [], {'primary_key': 'True'}),
+            'last_api_update': ('django.db.models.fields.DateTimeField', [], {'null': 'True', 'blank': 'True'}),
+            'last_update': ('django.db.models.fields.DateTimeField', [], {'auto_now': 'True', 'blank': 'True'}),
+            'subdomain': ('django.db.models.fields.CharField', [], {'max_length': '256'}),
+            'update_secret': ('django.db.models.fields.CharField', [], {'max_length': '256'})
+        }
+    }
+
+    complete_apps = ['main']

nsupdate/main/models.py

@@ -52,7 +52,11 @@ class Domain(models.Model):
     nameserver_update_key = models.CharField(max_length=256)
     nameserver_update_algorithm = models.CharField(
         max_length=256, default='HMAC_SHA512', choices=UPDATE_ALGORITHMS)
+    # XXX rename available_for_everyone to public
     available_for_everyone = models.BooleanField(default=False)
+    # available means "nameserver for domain operating and reachable" -
+    # gets set to False if we have trouble reaching the nameserver
+    available = models.BooleanField(default=True)
 
     last_update = models.DateTimeField(auto_now=True)
     created = models.DateTimeField(auto_now_add=True)
@@ -102,14 +106,16 @@ class Host(models.Model):
     def getIPv4(self):
         try:
             return dnstools.query_ns(self.get_fqdn(), 'A', origin=self.domain.domain)
-        except (dns.resolver.NXDOMAIN, dns.resolver.NoAnswer, dns.resolver.NoNameservers, dns.resolver.Timeout):
+        except (dns.resolver.NXDOMAIN, dns.resolver.NoAnswer, dns.resolver.NoNameservers, dns.resolver.Timeout,
-            return ''
+                dnstools.NameServerNotAvailable):
+            return 'error'
 
     def getIPv6(self):
         try:
             return dnstools.query_ns(self.get_fqdn(), 'AAAA', origin=self.domain.domain)
-        except (dns.resolver.NXDOMAIN, dns.resolver.NoAnswer, dns.resolver.NoNameservers, dns.resolver.Timeout):
+        except (dns.resolver.NXDOMAIN, dns.resolver.NoAnswer, dns.resolver.NoNameservers, dns.resolver.Timeout,
-            return ''
+                dnstools.NameServerNotAvailable):
+            return 'error'
 
     def poke(self):
         self.last_api_update = datetime.now()