Before proceeding, it is advised to check the release notes for your PDNS version, as specified in the name of the distribution file.
The 'it must still be Friday somewhere' release. Massive number of fixes, portability improvements and the new Geobackend by Mark Bergsma & friends.
New:
The Geobackend which makes it possible to send different answers to different IP ranges. Initial documentation can be found in pdns/modules/geobackend/README.
qgen query generation tool. Nearly completely undocumented and hard to build too, it requires Boost. But very spiffy. Use cd pdns; make qgen to build it.
Bugfixes:
The most reported bug ever was fixed. Zone2sql required the inclusion of unistd.h, except on Debian unstable.
PowerDNS tried to listen on its control "pipe" which does not work. Probably harmless, but might have caused some oddities.
The Packet Cache did not always set its TTL immediately, causing some packets to be inserted, even when running with the cache disabled (Mark Bergsma).
Valgrind found some unitialized reads, causing bogus values in the priority field when it was not needed
Valgrind found a bug in MTasker where we used delete instead of delete[].
SOA serials and other parameters are unsigned. This means that very large SOA serial numbers would be messed up (Michel Stol, Stefano Straus)
PowerDNS left its controlsocket around after exit and reported confusing errors if a socket was already in use.
The recursor proxy did not work on big endian systems like SPARC and some MIPS processors (Remco Post)
We no longer dump core on processing LOC records on UltraSPARC (Andrew Mulholland supplied a testing machine)
Improvements:
MySQL can now connect to a specified port again (Chris Anderton)
When running chroot()ed and with master or slave support active, PowerDNS needs to resolve domain names to find slaves. This in turn may require access to certain libraries. Previously, these needed to be available in the chroot directory but by forcing an initial lookup, these libraries are now loaded before the chrooting.
pdns_recursor was very slow after having done a larger number of queries because of the checks to see if a query should be throttled. This is now done using a set which is a lot faster than the previous full sequential scan.
The throttling code may not have throttled as much as was configured.
Yet another big LDAP update. The LDAP backend now loadbalances connections over several hosts (Norbert Sendetzky)
Updated b.root-servers.net address in the recursor
This release fixes up some of the shortcomings in 2.9.14, and adds some new features too.
Bugfixes:
allow-recursion-override was on by default, it was meant to be off.
Logging was still off in daemon mode, fixed.
debian/rules forgot to build an sqllite package
Recursor accidentally linked in MySQL - this was the result of an experiment with a persistent recursor cache.
The PowerDNS recursor had stability problems. It now sorts nameservers (roughly) by responsiveness. The 'roughly' part upset the sorting algorithm used, the speeds being sorted on changed during sorting.
The recursor now outputs the nameserver average response times in trace mode
LDAP compiles again.
Improvements:
zone2sql can now accept - as a filename which causes it to read stdin. This allows the following to work: dig axfr ds9a.nl | zone2sql --gmysql --zone=- | mysql pdns, which is a nice way to import a zone.
zone2sql now ignores duplicate SOA records which are identical - which also makes the above possible.
Remove libpqpp dependencies - since we now use the native C API for PostgreSQL
Big release with the fix for the all important 2^30 seconds problem and a lot of other news.
errno problems would cause compilation problems when using LDAP (Norbert Sendetzky)
The Generic SQL backend could cause crashes on PostgreSQL when using pdns_control notify (Georg Bauer)
Debian compatible init.d script (Wichert Akkerman)
If using the master or slave features, pdns had the notion of eternity ending in 2038, except that due to a thinko, eternity ended out to be the 10th of January 2004. This caused a loop to timeout immediately. Many thanks to Jasper Spaans for spotting the bug within five minutes.
Parts of the SOA field were not cannonicalized
The loglevel could in fact cause nothing to be logged (Norbert Sendetzky)
Improvements:
The recursor now chooses the fastest nameserver, which causes a big speedup!
LDAP now has different lookup models
Cleanups, better load distribution, better exception handling, zone2ldap improvements
The recursor was somewhat chatty about TCP connections
PostgreSQL now only depends on the C API and not on the deprecated C++ one
PowerDNS can now fully overrule external zones when doing recursion. See Chapter 11.
Big news! Windows is back! Our great friend Michel Stol found the time to update the PowerDNS code so it works again under windows.
Furthermore, big thanks go out to Dell who quickly repaired my trusty laptop.
His changes:
Generic SQLite support added
Removed the ODBC backend, replaced it by the Generic ODBC Backend, which has all the cool configurability of the Generic MySQL and PostgreSQL backends.
The PowerDNS Recursor now runs as a Service. It defaults to running on port 5300, PowerDNS itself is configured to expect the Recursor on port 5300 now.
The PowerDNS Service is now known as 'PowerDNS' to Windows.
The Installer was redone, this time with NSIS2.
General updates and fixes.
Other news:
![]() | There appears to be a problem with PowerDNS on Red Hat 7.3 with GCC 2.96 and self-compiled binaries. The symptoms are that PowerDNS works on the foreground but fails as a daemon. We're working on it. If you do note problems, let the list know, if you don't, please do so as well. Tell us if you use the RPM or compiled yourself. It is known that not compiling in MySQL support helps solve the problem, but then you don't have MySQL. |
There have been a number of reports on MySQL connections being dropped on FreeBSD 4.x, which sometimes causes PowerDNS to give up and reload itself. To combat this, MySQL error messages have been improved in some places in hopes of figuring out what is up. The initial indication is that MySQL itself sometimes terminates the connection and, amazingly, that switching to a Unix domain socket instead of TCP solves the problem.
Bug fixes:
allow-axfr-ips did not work for individual IP addresses (bug & fix by Norbert Sendetzky)
Improvements:
Opteron support! Thanks to Jeff Davey for providing a shell on an Opteron. The fixes should also help PowerDNS on other platforms with a 64 bit userspace.
Btw, the PowerDNS team has a strong desire for an Opteron :-)
pdns_recursor jumbles answers now. This means that you can do poor man's roundrobin by supplying multiple A, MX or AAAA records for a service, and get a random one on top each time. Interestingly, this feature appeared out of nowhere, this change was made to the authoritative code but due to the wonders of code-reuse had an effect on pdns_recursor too.
Big LDAP cleanup. Support for TLS was added. Zone2LDAP also gained the ability to generate ldif files containing a tree or a list of entries. (Norbert Sendetzky)
Zone2sql is now somewhat clearer when reporting malformed line errors - it did not always include the name of the file causing a problem, especially for big installations. Problem noted by Thom May.
pdns_recursor now survives the expiration of all its root records, most often caused by prolonged disconnection from the net.
Release rich in features. Work on Verisign oddities, addition of SQLite backend, pdns_recursor maturity.
New features:
--version command (requested by Mike Benoit)
delegation-only, a Verisign special. See Section 12.1.1.
Generic SQLite support, by Michel 'Who da man?' Stol. See Section A.7.
init.d script for pdns_recursor
Recursor now actually purges its cache, saving memory.
Slave configuration now no longer falls over when presented with a NULL master
Bindbackend2 now has supermaster support (Mark Bergsma, untested)
Answers are now shuffled! It turns out a few recursors don't do shuffling (pdns_recursor, djbdns), so we do it now. Requested by Jorn Ekkelenkamp of ISP-Services. This means that if you have multiple IP addresses for one host, they will be returned in differing order every once in a while.
Bugs:
0.0.0.0/0 didn't use to work (Norbert Sendetzky)
pdns_recursor would try to resolve IP address which to bind to, potentially causing chicken/egg problem
gpgsql no longer reports as gmysql (Sherwin Daganoto)
SRV would not be parsed right from disk (Christof Meerwald)
An AXFR from a zone hosted on the LDAP backend no longer transmits all the reverse entries too (Norbert Sendetzky)
PostgreSQL backend now does error checking. It would be a bit too trusting before.
Improvements, cleanups:
PowerDNS now reports the numerical IP addresses it binds to instead of the, possibly, alphanumeric names the operator passed.
Removed only-soa hackery (noticed by Norbert Sendetzky)
Debian packaging fixes (Wichert Akkerman)
Some parameter descriptions were improved.
Cleanups by Norbert: getAuth moved to chopOff, arguments::contains massive cleanup, more.
Yet another iteration, hopefully this will be the last silly release.
![]() | There has been a change in behaviour whereby disable-axfr does what it means now! From now on, setting allow-axfr-ips automatically disables AXFR from unmentioned subnets. |
This release enables AXFR again, disable-axfr did the opposite of what it claimed. Furthermore, the pdns_recursor now cleans its cache, which should save some memory in the long run. Norbert contributed some small LDAP work which should come in useful in the future.
Small bugfixes, LDAP update. Released 3rd of July 2003. Apologies for the long delay, real life keeps interfering.
![]() | Do not use or try to use 2.9.9, it was a botched release! |
![]() | There has been a change in behaviour whereby disable-axfr does what it means now! From now on, setting allow-axfr-ips automatically disables AXFR from unmentioned subnets. |
2.9.8 was prone to crash on adding additional records. Thanks to excellent debugging by PowerDNS users worldwide, the bug was found quickly and is in fact present in all earlier PowerDNS releases, but for some reason doesn't cause crashes there.
Notifications now jump in front of the queue of domains that need to be checked for changes, giving much greater perceived performance. This is needed if you have tens of thousands of slave domains and your master server is on a high latency link. Thanks to Mark Jeftovic of EasyDNS for suggesting this change and testing it on their platform.
Dean Mills reported that PowerDNS does confusing logging about changing GIDs and UIDs, fixed. Cosmetic only.
pdns_recursor may have logged empty lines for some users, fixed. Solution suggested by Norbert Sendetzky.
LDAP: DNS TTLs were random values (Norbert Sendetzky, Stefan Pfetzing). New ldap-default-ttl option.
LDAP: Now works with OpenLDAP 2.1 (Norbert Sendetzky)
LDAP: error handling for invalid MX records implemented (Norbert Sendetzky)
LDAP: better exception handling (Norbert Sendetzky)
LDAP: code cleanup of lookup() (Norbert Sendetzky)
LDAP: added support for scoped searches (Norbert Sendetzky)
Queen's day release! 30th of April 2003.
Added support for AIX, fixed negative SOA caching. Some other cleanups. Not a major release but enough reasons to upgrade.
Bugs fixed:
Recursor had problems expiring negatively cached entries, which wasted memory and also led to the continued non-existence of hosts that since had come into existence.
The Generic SQL backends did not lowercase the names of records, which led to new records not being found by case sensitive databases (notably PostgreSQL). Found by Volker Goetz.
NS queries for zones for which we did not carry authority, but only had delegation information, had their NS records in the wrong section. Minor detail, but a standards violation on etheless. Spotted by Stephane Bortzmeyer.
Improvements:
Removed crypt.h dependency from powerldap.hh, which was a problem on some platforms (Richard Arends)
PowerDNS can't parse so called binary labels which we now detect and ignore, after printing a warning.
Specifying allow-axfr-ips now automatically disables AXFR for all non-mentioned addresses.
A Solaris ready init.d script is now part of the tar.gz (contributed, but I lost by whom).
Added some fixes to PowerDNS can work on AIX (spotted by Markus Heimhilcher).
Norbert Sendetzky contributed zone2ldap.
Everybody's favorite compiler warning from zone2sql.cc was removed!
Recursor now listens on TCP!
Released on 2003-03-20.
This is a sweeping release in the sense of cleanup. There are some new features but mostly a lot of cleanup going on. Hiding inside is the bind2backend, the next generation of the bind backend. A work in progress. Those of you with overlapping zones, as mentioned in the changelog of 2.9.6, are invited to check it out by replacing launch=bind by launch=bind2 and renaming all bind- parameters to bind2-. Be aware that if you run with many small zones, this backend is faster, but if you run with a few large ones, it is slower. This will improve.
Features:
Mark Bergsma contributed query-local-address which allows the operator to select which source address to use. This is useful on servers with multiple source addresses and the operating system selecting an unintended one, leading to remotes denying access.
PowerDNS can now perform AAAA additional processing optionally, turned on by setting do-ipv6-additional-processing. Thanks to Stephane Bortzmeyer for pointing out the need.
Bind2backend, which is almost in compliance with the new IETF AXFR-clarify (some would say 'redefinition') draft.
This backend is not ready for primetime but you may want to try it if you currently have overlapping zones and note problems. An overlapping zone would be having "ipv6.powerdns.com" and "powerdns.com" zones on one server.
Improvements:
Zone2sql would happily try to read from a directory and not give a useful error about this.
PowerDNS now reports the case where it can't figure out any IP address of slave nameservers for a zone
Removed receiver-threads setting which was experimental and in fact only made things worse.
LDAP backend updates from its author Norbert Sendetzky. Reverse lookups should work now too.
An error message about unparseable packets did not include the originating IP address (fixed by Mark Bergsma)
PowerDNS can now be started via path resolution while running with a guardian. Suggested by Maurice Nonnekes.
pdns_recursor moved to sbin (reported by Norbert Sendetzky)
Retuned some logger errorlevels, a lot of master/slave chatter was logged as 'Error'. Reported by Willem de Groot.
Bugs fixed:
zone2sql did not remove trailing dots in SOA records.
ldapbackend did not include utility.hh which caused compilation problems on Solaris (reported by Remco Post)
pdns_control could leave behind remnants in case PowerDNS was not running (reported by dG)
Incoming AXFR did not work on Solaris and other big-endian systems (Willem de Groot helped debugging this long standing problem).
Recursor could crash on convoluted CNAME loops. Thanks to Dan Faerch for delivering coredumps.
Silly 'wuh' debugging output in zone2sql and bindbackend removed (spotted by Ivo van der Wijk)
Recursor neglected to differentiate between negative cache of NXDOMAIN and NOERROR, leading to problems with IPv6 enabled Windows clients. Thanks to Stuart Walsh for reporting this and testing the fix.
PowerDNS set the 'aa' bit on serving NS records in a zone for which it was authoritative. Most implementations drop the 'aa' bit in this case and Stephane Bortzmeyer informed us of this. PowerDNS now also drops the 'aa' bit in this case.
The webserver tended to fail after prolonged operation on FreeBSD, this was due to an uninitialised timeout, other platforms were lucky. Thanks to G.P. de Boer for helping debug this.
getAnswers() in dnspacket.cc could be forced to read bytes beyond the end of the packet, leading to crashes in the PowerDNS recursor. This is an ongoing project that needs more work. Reported by Dan Faerch, with a coredump proving the problem.
Two new backends - Generic ODBC (windows only) and LDAP. Furthermore, a few important bugs have been fixed which may have hampered sites seeing a lot of outgoing zonetransfers. Additionally, the pdns recursor now has 'query throttling' which is pretty cool. In short this makes sure that PowerDNS does not send out heaps of queries if a nameserver is unable to provide an answer. Many operators of authoritative setups are all too aware of recursing nameservers that hammer them for zones they don't have, PowerDNS won't do that anymore now, no matter what clients request of it.
![]() | There is an unresolved issue with the BIND backend and 'overlapping' slave zones. So if you have 'example.com' and also have a separate slave zone called 'external.example.com', things may go wrong badly. Thanks to Christian Laursen for working with us a lot in finding this issue. We hope to resolve it soon. |
BIND Backend now honours notifies, code to support this was accidentally left out. Thanks to Christian Laursen for noticing this.
Massive speedup for those of you using the slightly deprecated MBOXFW records. Thanks to Jorn of ISP Services for helping and testing this improvement.
$GENERATE had an off-by-one bug where it would omit the last record to be generated (Christian Laursen)
Simultaneous AXFRs may have been problematic on some backends. Thanks to Jorn of ISP-Services again for helping us resolve this issue.
Added LDAP backend by Norbert Sendetzky, see Section A.12.
Added Generic ODBC backend for Windows by Michel Stol.
Simplified 'out of zone data' detection in incoming AXFR support, hopefully removing a case sensitivity bug there. Thanks again to Christian Laursen for reporting this issue.
$include in-zonefile was broken under some circumstances, losing the last character of a filename. Thanks to Joris Vandalon for noticing this.
The zoneparser was more case-sensitive than BIND, refusing to accept 'in' as well as 'IN'. Thanks to Joris Vandalon for noticing this.
Released on 2002-02-03.
This version is almost entirely about recursion with major changes to both the pdns recursor, which is renamed to 'pdns_recursor' and to the main PowerDNS binary to make it interact better with the recursing component.
Sadly, due to technical reasons, compiling the pdns recursor and pdns authoritative nameserver into one binary is not immediately possible. During the release of 2.9.4 we stated that the recursing nameserver would be integrated in the next release - this won't happen now.
However, this turns out to not be that bad at all. The recursor can now be restarted without having to restart the rest of the nameserver, for example. Cooperation between the both halves of PDNS is also almost seamless. As a result, 'non-lazy recursion' has been dropped. See Chapter 11 for more details.
Furthermore, the recursor only works on Linux, Windows and Solaris (not entirely). FreeBSD does not support the required functions. If you know any important FreeBSD people, plea with them to support set/get/swapcontext! Alternatively, FreeBSD coders could read the solution presented here in figure 5.
The 'Contributor of the Month' award goes to Mark Bergsma who has responded to our plea for help with the label compressor and contributed a wonderfully simple and right fix that allows PDNS to compress just as well as Other namerervers out there. An honorary mention goes to Ueli Heuer who, despite having no C++ experience, submitted an excellent SRV record implementation.
Excellent work was also performed by Michel Stol, the Windows guy, in fixing all our non-portable stuff again. Christof Meerwald has also done wonderful work in porting MTasker to Windows, which was then used by Michel to get the recursor functioning on Windows.
Other changes:
dnspacket.cc was cleaned up by factoring out common operations
Heaps of work on the recursing nameserver. Has now achieved *days* of uptime!
Recursor renamed from syncres to pdns_recursor
PowerDNS can now serve records it does not know about. To benefit from this slightly undocumented feature, add 1024 to the numerical type of a record and include the record in binary form in your database. Used internally by the recursing nameserver but you can use it too.
PowerDNS now knows about SIG and KEY records *names*. It does not support them yet but can at least report so now.
HINFO records can now be transferred from a master to PowerDNS (thanks to Ueli Heuer for noticing it didn't work).
Yet more UltraSPARC alignment issues fixed (Chris Andrews).
Dropped non-lazy recursion, nobody was using it. Lazy recursion became even more lazy after Dan Bernstein pointed out that additional processing is not vital, so PowerDNS does its best to do additional processing on recursive queries, but does not scream murder if it does not succeed. Due to caching, the next identical query will be successfully additionally processed.
Label compression was improved so we can now fit all . records in 436 bytes, this used to be 460! (Code & formal proof of correctness by Mark Bergsma).
SRV support (incoming and outgoing), submitted by Ueli Heuer.
Generic backends do not support SOA serial autocalculation, it appears. Could lead to random SOA serials in case of a serial of 0 in the database. Fixed so that 0 stays zero in that case. Don't set the SOA serial to 0 when using Generic MySQL or Generic PostgreSQL!
J root-server address was updated to its new location.
SIGUSR1 now forces the recursor to print out statistics to the log.
Meaning of recursor logging was changed a bit - a cache hit is now a question that was answered with 0 outgoing packets needed. Used to be a weighted average of internal cache hits.
MySQL compilation did not include -lz which causes problems on some platforms. Thanks to James H. Cloos Jr for reporting this.
After a suggestion by Daniel Meyer and Florus Both, the built in webserver now reports the configuration name when multiple PowerDNS instances are active.
Brad Knowles noticed that zone2sql had problems with the root.zone, fixed. This also closes some other zone2sql annoyances with converting single zones.
Yet another grand release. Big news is the addition of a recursing nameserver which has sprung into existence over the past week. It is in use on several computers already but it is not ready for prime time. Complete integration with PowerDNS is expected around 2.9.5, for now the recursor is a separate program.
In preliminary tests, the recursor appears to be four times faster than BIND 9 on a naive benchmark starting from a cold cache. BIND 9 managed to get through to some slower nameservers however, which were given up on by PowerDNS. We will continue to tune the recursor. See Chapter 12 for further details.
The BIND Backend has also been tested (see the bind-domain-status item below) rather heavily by several parties. After some discussion online, one of the BIND authors ventured that the newsgroup comp.protocols.dns.bind may now in fact be an appropriate venue for discussing PowerDNS. Since this discussion, traffic to the PowerDNS pages has increased sixfold and shows no signs of slowing down.
From this, it is apparent that far more people are interested in PowerDNS than yet know about it. So spread the word!
In other news, we now have a security page at Section 1.4. Furthermore, Maurice Nonnekes contributed an OpenBSD port! See his page for more details!
New features and improvements:
All SQL queries in the generic backends are now available for configuration. (Martin Klebermass/bert hubert). See Section A.5.
A recursing nameserver! See Chapter 12.
An incoming AXFR now only starts a backend zone replacement transaction after the first record arrived successfully, thus making sure no work is done when a remote nameserver is unable/unwilling to AXFR a zone to us.
Zoneparser error messages were improved slightly (thanks to Stef van Dessel for spotting this shortcoming)
XS4ALL's Erik Bos checked how PowerDNS reacted to a BIND installation with almost 60.000 domains, some of which with >100.000 records, and he discovered the pdns_control bind-domain-status command became very slow with larger numbers of domains. Fixed, 60.000 domains are now listed in under one second.
If a remote nameserver disconnects during an incoming AXFR, the update is now rolled back, unless the AXFR was properly terminated.
The migration chapter mentioned the use of deprecated backends.
A tremendous number of bugs were discovered and fixed:
Zone parser would only accept $include and not $INCLUDE
Zone parser had problems with $lines with comments on the end
Wildcard ANY queries were broken (thanks Colemarcus for spotting this)
A connection failure with the Generic backends would lead to a powerdns reload (cast of many)
Generic backends had some semantic problems with slave support. Symptoms were oft-repeated notifications and transfers (thanks to Mark Bergsma for helping resolve this).
Solaris version compiles again. Thanks to Mohamed Lrhazi for reporting that it didn't.
Some UltraSPARC alignment fixes. Thanks to Mohamed Lrhazi for being helpful in spotting these. One problem is still outstanding, Mohamed sent a core dump that tells us where the problem is. Expect the fix to be in 2.9.5. Volunteers can grep the source for 'UltraSPARC' to find where the problem is.
Our support of IPv6 on FreeBSD had phase of moon dependent bugs, fixed by Peter van Dijk.
Some crashes of and by pdns_control were fixed, thanks to Mark Bergsma for helping resolve these.
Outgoing AXFR in pdns installations with multiple loaded backends was broken (thanks to Stuart Walsh for reporting this).
A failed BIND Backend incoming AXFR would block the zone until it succeeded again.
Generic PostgreSQL backend wouldn't compile with newer libpq++, fixed by Julien Lemoine/SpeedBlue.
Potential bug (not observed) when listening on multiple interfaces fixed.
Some typos in manpages fixed (reported by Marco Davids).
![]() | 2.9.3a is identical to 2.9.3 except that zone2sql does work |
Broad range of huge improvements. We now have an all-static .rpm and .deb for Linux users and a a link to an OpenBSD port. Major news is that work on the Bind backend has progressed to the point that we've just retired our last Bind server and replaced it with PowerDNS in Bind mode! This server is operating a number of master and slave setups so it should stress the Bind backend somewhat.
This version is rapidly approaching the point where it is a better-Bind-than-Bind and nearly a drop-in replacement for authoritative setups. PowerDNS is now equipped with a powerful master/slave apparatus that offers a lot of insight and control to the user, even when operating from Bind zonefiles and a Bind configuration. Observe.
After the SOA of ds9a.nl was raised:
pdns[17495]: All slave domains are fresh pdns[17495]: 1 domain for which we are master needs notifications pdns[17495]: Queued notification of domain 'ds9a.nl' to 195.193.163.3 pdns[17495]: Queued notification of domain 'ds9a.nl' to 213.156.2.1 pdns[17520]: AXFR of domain 'ds9a.nl' initiated by 195.193.163.3 pdns[17520]: AXFR of domain 'ds9a.nl' to 195.193.163.3 finished pdns[17521]: AXFR of domain 'ds9a.nl' initiated by 213.156.2.1 pdns[17521]: AXFR of domain 'ds9a.nl' to 213.156.2.1 finished pdns[17495]: Removed from notification list: 'ds9a.nl' to 195.193.163.3 (was acknowledged) pdns[17495]: Removed from notification list: 'ds9a.nl' to 213.156.2.1 (was acknowledged) pdns[17495]: No master domains need notificationsIf however our slaves would ignore us, as some are prone to do, we can send some additional notifications:
$ sudo pdns_control notify ds9a.nl Added to queue pdns[17492]: Notification request for domain 'ds9a.nl' received pdns[17492]: Queued notification of domain 'ds9a.nl' to 195.193.163.3 pdns[17492]: Queued notification of domain 'ds9a.nl' to 213.156.2.1 pdns[17495]: Removed from notification list: 'ds9a.nl' to 195.193.163.3 (was acknowledged) pdns[17495]: Removed from notification list: 'ds9a.nl' to 213.156.2.1 (was acknowledged)Conversely, if PowerDNS needs to be reminded to retrieve a zone from a master, a command is provided:
$ sudo pdns_control retrieve forfun.net Added retrieval request for 'forfun.net' from master 212.187.98.67 pdns[17495]: AXFR started for 'forfun.net', transaction started pdns[17495]: Zone 'forfun.net' (/var/cache/bind/forfun.net) reloaded pdns[17495]: AXFR done for 'forfun.net', zone committedAlso, you can force PowerDNS to reload a zone from disk immediately with pdns_control bind-reload-now. All this happens 'live', per your instructions. Without instructions, the right things also happen, but the operator is in charge.
For more about all this coolness, see Section B.1.1 and Section A.9.2.
![]() | Again some changes in compilation instructions. The hybrid pgmysql backend has been split up into 'gmysql' and 'gpgsql', sharing a common base within the PowerDNS server itself. This means that you can no longer compile --with-modules="pgmysql" --enable-mysql --enable-pgsql but that you should now use: --with-modules="gmysql gpgsql". The old launch-names remain available. If you launch the Generic PgSQL backend as gpgsql2, all parameters will have gpsql2 as a prefix, for example pgsql2-dbname. If launched as gpsql, the regular names are in effect. |
![]() | The pdns_control protocol was changed which means that older pdns_controls cannot talk to 2.9.3. The other way around is broken too. This may lead to problems with automatic upgrade scripts, so pay attention if your daemon is truly restarted. Also make sure no old pdns_control command is around to confuse things. |
Improvements:
Bind backend can now deal with missing files and try to find them later.
Bind backend is now explicitly master capable and triggers the sending of notifications.
General robustness improvements in Bind backend - many errors are now non-fatal.
Accessability, Serviceability. New pdns_server commands like bind-list-rejects (lists zones that could not be loaded, and the reason why), bind-reload-now (reload a zone from disk NOW), rediscover (reread named.conf NOW). More is coming up.
Added support for retrieving RP (Responsible Person) records from remote masters. Serving them was already possible.
Added support for LOC records, which encode the geographical location of a host, both serving and retrieving (thanks to Marco Davids using them on our last Bind server, forcing us to implement this silly record).
Configuration file parser now strips leading spaces too, allowing "chroot= /tmp" to work, as well as "chroot=/tmp" (Thanks to Hub Dohmen for reporting this for months on end).
Added bind-domain-status command that shows the status of all domains (when/if they were parsed, any errors encountered while parsing them).
Added bind-reload-now command that tries to reload a zone from disk NOW, and reports back errors to the operator immediatly.
Added retrieve command that queues a request to retrieve a zone from its master.
Zones retrieved from masters are now stored way smaller on disk because the domain is stripped from records, which is derived from the configuration file. Retrieved zones are now prefixed with some information on where they came from.
Changes:
gpgsql and gmysql backends split out of the hybrid pgmysqlbackend. This again changed compilation instructions!
pdns_control now uses the rarely seen SOCK_STREAM Unix Domain socket variety so it can transport large amounts of text, which is needed for the bind-domain-status command, for which see Section A.9.2. This breaks compatability with older pdns_control and pdns_server binaries!
Bind backend now ignores 'hint' and 'forward' and other unsupported zone types.
AXFRs are now logged more heavily by default. An AXFR is a heavy operation anyhow, some more logging does not further increase the load materially. Does help in clearing up what slaves are doing.
A lot of master/slave chatter has been silenced, making output more relevant. No more repetitive 'No master domains need notifications' etc, only changes are reported now.
Bugfixes:
Windows version did not compile without minor changes.
Confusing error reporting on Windows 98 (which does not support PowerDNS) fixed
Potential crashes with shortened packets addressed. An upgrade is advised!
notify (which was already there, just badly documented) no longer prints out debugging garbage.
pgmysql backend had problems launching when not compiled in but available as a module. Workaround for 2.9.2 is 'load-modules=pgmysql', but even then gpgsql would not work! gmysql would then, however. These modules are now split out, removing such issues.
Bugfixes galore. Solaris porting created some issues on all platforms. Great news is that PowerDNS is now in Debian 'sid' (unstable). The 2.9.1 packages in there currently aren't very good but the 2.9.2 ones will be. Many thanks to Wichert Akkerman, our 'downstream' for making this possible.
![]() | The Generic MySQL backend, part of the Generic MySQL & PostgreSQL backend, is now the DEFAULT! The previous default, the 'mysql' backend (note the lack of 'g') is now DEPRECATED. This was the source of much confusion. The 'mysql' backend does not support MASTER or SLAVE operation. The Generic backends do. To get back the mysql backend, add --with-modules="mysql" or --with-dynmodules="mysql" if you prefer to load your modules at runtime. |
Bugs fixed:
Silly debugging output removed from the webserver (found by Paul Wouters)
SEVERE: due to Solaris portability fixes, qtypes<127 were broken. These include NAPTR, ANY and AXFR. The upshot is that powerdns wasn't performing outgoing AXFRs nor ANY queries. These were the 'question for type -1' warnings in the log
incoming AXFR could theoretically miss some trailing records (not observed, but could happen)
incoming AXFR did not support TXT records (spotted by Paul Wouters)
with some remotes, an incoming AXFR would not terminate until a timeout occured (observed by Paul Wouters)
Documentation bug, pgmysql != mypgsql
Documentation:
Documented the 'random backend', see Section A.3.
Wichert Akkerman contributed three manpages.
Building PowerDNS on Unix is now documented somewhat more, see Section D.1.
Features:
pdns init.d script is now +x by default
OpenBSD is on its way of becoming a supported platform! As of 2.9.2, PowerDNS compiles on OpenBSD but swiftly crashes. Help is welcome.
ODBC backend (for Windows only) was missing from the distribution, now added.
xdb backend added - see Section A.11. Designed for use by root-server operators.
Dynamic modules are back which is good news for distributors who want to make a pdns packages that does not depend one every database under the sun.
Thanks to the great enthusiasm from around the world, powerdns is now available for Solaris and FreeBSD users again! Furthermore, the Windows build is back. We are very grateful for the help of:
Michel Stol
Wichert Akkerman
Edvard Tuinder
Koos van den Hout
Niels Bakker
Erik Bos
Alex Bleker
steven stillaway
Roel van der Made
Steven Van Steen
We are happy to have been able to work with the open source community to improve PowerDNS!
Changes:
The monitor command set no longer allows the changing of non-existant variables.
IBM Universal Database DB2 backend now included in source distribution (untested!)
Oracle backend now included in source distribution (sligthly tested!)
configure script now searches for postgresql and mysql includes
Bind parser now no longer dies on records with a ' in them (Erik Bos)
The pipebackend was accidentally left out of 2.9
FreeBSD fixes (with help from Erik Bos, Alex Bleeker, Niels Bakker)
Heap of Solaris work (with help from Edvard Tuinder, Stefan Van Steen, Koos van den Hout, Roel van der Made and especially Mark Bakker). Now compiles in 2.7 and 2.8, haven't tried 2.9. May be a bit dysfunctional on 2.7 though - it won't do IPv6 and it won't serve AAAA. Patches welcome!
Windows 32 build is back! Michel Stol updated his earlier work to the current version.
S/Linux (Linux on Sparc) build works now (with help from steven stillaway).
Silly debugging message ('sd.ttl from cache') removed
.debs are back, hopefully in 'sid' soon! (Wichert Akkerman)
Removal of bzero and other less portable constructs. Discovered that recent Linux glibc's need -D_GNU_SOURCE (Wichert Akkerman).
Open source release. Do not deploy unless you know what you are doing. Stability is expected to return with 2.9.1, as are the binary builds.
License changed to the GNU General Public License version 2.
Cleanups by Erik Bos @ xs4all.
Build improvements by Wichert Akkerman
Lots of work on the build system, entirely revamped. By PowerDNS.
From this release onwards, we'll concentrate on stabilising for the 3.0 release. So if you have any must-have features, let us know soonest. The 2.8 release fixes a bunch of small stability issues and add two new features. In the spirit of the move to stability, this release has already been running 24 hours on our servers before release.
pipe backend gains the ability to restricts its invocation to a limited number of requests. This allows a very busy nameserver to still serve packets from a slow perl backend.
pipe backend now honors query-logging, which also documents which queries were blocked by the regex.
pipe backend now has its own backend chapter.
An incoming AXFR timeout at the wrong moment had the ability to crash the binary, forcing a reload. Thanks to our bug spotting champions Mike Benoit and Simon Kirby of NetNation for reporting this.
This version fixes some very long standing issues and adds a few new features. If you are still running 2.6, upgrade yesterday. If you were running 2.6.1, an upgrade is still strongly advised.
Features:
The controlsocket is now readable and writable by the 'setgid' user. This allows for non-root access to PDNS which is nice for mrtg or cricket graphs.
MySQL backend (the non-generic one) gains the ability to read from a different table using the mysql-table setting.
pipe backend now has a configurable timeout using the pipe-timeout setting. Thanks fo Steve Bromwich for pointing out the need for this.
Experimental backtraces. If PowerDNS crashes, it will log a lot of numbers and sometimes more to the syslog. If you see these, please report them to us. Only available under Linux.
Bugs:
2.7 briefly broke the mysql backend, so don't use it if you use that. 2.7.1 fixes this.
SOA records could sometimes have the wrong TTL. Thanks to Jonas Daugaard for reporting this.
An ANY query might lead to duplicate SOA records being returned under exceptional circumstances. Thanks to Jonas Daugaard for reporting this.
Underlying the above bug, packet compression could sometimes suddenly be turned off, leading to overly large responses and non-removal of duplicate records.
The allow-axfr-ips setting did not accept IP ranges (1.2.3.0/24) which the documentation claimed it did (thanks to Florus Both of Ascio technologies for being sufficiently persistent in reporting this).
Killed backends were not being respawned, leading to suboptimal behaviour on intermittent database errors. Thanks to Steve Bromwich for reporting this.
Corrupt packets during an incoming AXFR when acting as a slave would cause a PowerDNS reload instead of just failing that AXFR. Thanks to Mike Benoit and Simon Kirby of NetNation for reporting this.
Label compression in incoming AXFR had problems with large offsets, causing the above mentioned errors. Thanks to Mike Benoit and Simon Kirby of NetNation for reporting this.
Quick fix release for a big cache problem.
Performance release. A lot of work has been done to raise PDNS performance to staggering levels in order to take part in benchmarketing efforts. Together with our as yet unnamed partner, PDNS has been benchmarked at 60.000 mostly cached queries/second on off the shelf PC hardware. Uncached performance was 17.000 uncached DNS queries/second on the .ORG domain.
Performance has been increased by both making PDNS itself quicker but also by lowering the number of backend queries typically needed. Operators will typically see PDNS taking less CPU and the backend seeing less load.
Furthermore, some real bugs were fixed. A couple of undocumented performance switches may appear in --help output but you are advised to stay away from these.
Developers: this version needs the pdns-2.5.1 development kit, available on http://downloads.powerdns.com/releases/dev. See also Appendix C.
Performance:
A big error in latency calculations - cached packets were weighed 50 times less, leading to inflated latency reporting. Latency calculations are now correct and way lower - often in the microseconds range.
It is now possible to run with 0 second cache TTLs. This used to cause very frequent cache cleanups, leading to performance degradation.
Many tiny performance improvements, removing duplicate cache key calculations, etc. The cache itself has also been reworked to be more efficient.
First 'CNAME' backend query replaced by an 'ANY' query, which most of the time returns the actual record, preventing the need for a separate CNAME lookup, halving query load.
Much of the same for same-level-NS records on queries needing delegation.
Bugs fixed:
Incidentally, the cache count would show 'unknown' packets, which was harmless but confusing. Thanks to Mike and Simon of NetNation for reporting this.
SOA hostmaster with a . in the local-part would be cached wrongly, leading to a stray backslash in case of multiple successively SOA queries. Thanks to Ascio Techologies for spotting this bug.
zone2sql did not parse Verisign zonefiles correctly as these contained a $TTL statement in mid-record.
Sometimes packets would not be accounted, leading to 'udp-queries' and 'udp-answers' divergence.
Features:
'cricket' command added to init.d scripts that provides unadorned output for parsing by 'Cricket'.
Brown paper bag release fixing a huge memory leak in the new Query Cache.
Developers: this version needs the new pdns-2.5.1 development kit, available on http://downloads.powerdns.com/releases/dev. See also Appendix C.
And some small changes:
Added support for RFC2038 compliant negative-answer caching. This allows remotes to cache the fact that a domain does not exist and will not exist for a while. Thanks to Chris Thompson for pointing out how tiny our minds are. This feature may cause a noticeable reduction in query load.
Small speedup to non-packet-cached queries, incidentally fixing the huge memory leak.
pdns_control ccounts command outputs statistics on what is in the cache, which is useful to help optimize your caching strategy.
An important release which has seen quite a lot of trial and error testing. As a result, PDNS can now run with a huge cache and concurrent invalidations. This is useful when running of a slower database or under high traffic load with a fast database.
Furthermore, the gpgsql2 backend has been validated for use and will soon supplant the gpgsql backend entirely. This also bodes well for the gmysql backend which is the same code.
Also, a large amount of issues biting large scale slave operators were addressed. Most of these issues would only show up after prolonged uptime.
New features:
Query cache. The old Packet Cache only cached entire questions and their answers. This is very CPU efficient but does not lead to maximum hitrate. Two packets both needing to resolve smtp.you.com internally would not benefit from any caching. Furthermore, many different DNS queries lead to the same backend queries, like 'SOA for .COM?'.
PDNS now also caches backend queries, but only those having no answer (the majority) and those having one answer (almost the rest).
In tests, these additional caches appear to halve the database backend load numerically and perhaps even more in terms of CPU load. Often, queries with no answer are more expensive than those having one.
The default ttls for the query-cache and negquery-cache are set to safe values (20 and 60 seconds respectively), you should be seeing an improvement in behaviour without sacrificing a lot in terms of quick updates.
The webserver also displays the efficiency of the new Query Cache.
The old Packet Cache is still there (and useful) but see Chapter 9 for more details.
There is now the ability to shut off some logging at a very early stage. High performance sites doing thousands of queries/second may in fact spend most of their CPU time on attempting to write out logging, even though it is ignored by syslog. The new flag log-dns-details, on by default, allows the operator to kill most informative-only logging before it takes any cpu.
Flags which can be switched 'on' and 'off' can now also be set to 'off' instead of only to 'no' to turn them off.
Enhancements:
Packet Cache is now case insensitive, leading to a higher hitrate because identical queries only differing in case now both match. Care is taken to restore the proper case in the answer sent out.
Packet Cache stores packets more efficiently now, savings are estimated at 50%.
The Packet Cache is now asynchronous which means that PDNS continues to answer questions while the cache is busy being purged or queried. Incidentally this will mean a cache miss where previously the question would wait until the cache became available again.
The upshot of this is that operators can call pdns_control purge as often as desired without fearing performance loss. Especially the full, non-specific, purge was speeded up tremendously.
This optimization is of little merit for small sites but is very important when running with a large packetcache, such as when using recursion under high load.
AXFR log messages now all contain the word 'AXFR' to ease grepping.
Linux static version now compiled with gcc 3.2 which is known to output better and faster code than the previously used 3.0.4.
Bugs fixed:
Packetcache would sometimes send packets back with slightly modified flags if these differed from the flags of the cached copy.
Resolver code did bad things with filedescriptors leading to fd exhaustion after prolonged uptimes and many slave SOA currency checks.
Resolver code failed to properly log some errors, leading to operator uncertainty regarding to AXFR problems with remote masters.
After prolonged uptime, slave code would try to use privileged ports for originating queries, leading to bad replication efficiency.
Masters sending back answers in differing case from questions would lead to bogus 'Master tried to sneak in out-of-zone data' errors and failing AXFRs.
Developers: this version is compatible with the pdns-2.1 development kit, available on http://downloads.powerdns.com/releases/dev. See also Appendix C.
This version fixes some stability issues with malformed or malcrafted packets. An upgrade is advised. Furthermore, there are interesting new features.
New features:
Recursive queries are now also cached, but in a separate namespace so non-recursive queries don't get recursed answers and vice versa. This should mean way lower database load for sites running with the current default lazy-recursion. Up to now, each and every recursive query would lead to a large amount of SQL queries.
To prevent the packetcache from becoming huge, a separate recursive-cache-ttl can be specified.
The ability to change parameters at runtime was added. Currently, only the new query-logging flag can be changed.
Added query-logging flag which hints a backend that it should output a textual representation of queries it receives. Currently only gmysql and gpgsql2 honor this flag.
Gmysql backend can now also talk to PgSQL, leading to less code. Currently, the old postgresql driver ('gpgsql') is still the default, the new driver is available as 'gpgsql2' and has the benefit that it does query logging. In the future, gpgsql2 will become the default gpgsql driver.
DNS recursing proxy is now more verbose in logging odd events which may be caused by buggy recursing backends.
Webserver now displays peak queries/second 1 minute average.
Bugs fixed:
Failure to connect to database in master/slave communicator thread could lead to an unclean reload, fixed.
Documentation: added details for strict-rfc-axfrs. This feature can be used if very old clients need to be able to do zone transfers with PDNS. Very slow.
Developers: this version is compatible with the pdns-2.1 development kit, available on http://downloads.powerdns.com/releases/dev. See also Appendix C.
This release adds the Generic MySQL backend which allows full master/slave semantics with MySQL and InnoDB tables (or other tables that support transactions). See Section A.5.
Other new features:
Improved error messages in master/slave communicator will help down track problems.
slave-cycle-interval setting added. Very large sites with thousands of slave domains may need to raise this value above the default of 60. Every cycle, domains in undeterminate state are checked for their condition. Depending on the health of the masters, this may entail many SOA queries or attempted AXFRs.
Bugs fixed:
'pdns_control purge domain' and 'pdns_control purge domain$' were broken in version 2.2 and did not in fact purge the cache. There is a slight risk that domain-specific purge commands could force a reload in previous version. Thanks to Mike Benoit of NetNation for discovering this.
Master/slave communicator thread got confused in case of delayed answers from slow masters. While not causing harm, this caused inefficient behaviour when testing large amounts of slave domains because additional 'cycles' had to pass before all domains would have their status ascertained.
Backends implementing special SOA semantics (currently only the undocumented 'pdns express backend', or homegrown backends) would under some circumstances not answer the SOA record in case of an ANY query. This should put an end to the last DENIC problems. Thanks to DENIC for helping us find the problem.
Developers: this version is compatible with the pdns-2.1 development kit, available on http://downloads.powerdns.com/releases/dev. See also Appendix C.
Again a big release. PowerDNS is seeing some larger deployments in more demanding environments and these are helping shake out remaining issues, especially with recursing backends.
The big news is that wildcard CNAMEs are now supported, an oft requested feature and nearly the only part in which PDNS differed from BIND in authoritative capabilities.
If you were seeing signal 6 errors in PDNS causing reloads and intermittent service disruptions, please upgrade to this version.
For operators of PowerDNS Express trying to host .DE domains, the very special soa-serial-offset feature has been added to placate the new DENIC requirement that the SOA serial be at least six digits. PowerDNS Express uses the SOA serial as an actual serial and not to insert dates and hence often has single digit soa serial numbers, causing big problems with .DE redelegations.
Bugs fixed:
Malformed or shortened TCP recursion queries would cause a signal 6 and a reload. Same for EOF from the TCP recursing backend. Thanks to Simon Kirby and Mike Benoit of NetNation for helping debug this.
Timeouts on the TCP recursing backend were far too long, leading to possible exhaustion of TCP resolving threads.
pdns_control purge domain accidentally cleaned all packets with that name as a prefix. Thanks to Simon Kirby for spotting this.
Improved exception error logging - in some circumstances PDNS would not properly log the cause of an exception, which hampered problem resolution.
New features:
Wildcard CNAMEs now work as expected!
pdns_control purge can now also purge based on suffix, allowing operators to purge an entire domain from the packet cache instead of only specific records. See also Section B.1.1 Thanks to Mike Benoit for this suggestion.
soa-serial-offset for installations with small SOA serial numbers wishing to register .DE domains with DENIC which demands six-figure SOA serial numbers. See also Chapter 15.
This is a somewhat bigger release due to pressing demands from customers. An upgrade is advised for installations using Recursion. If you are using recursion, it is vital that you are aware of changes in semantics. Basically, local data will now override data in your recursing backend under most circumstances. Old behaviour can be restored by turning lazy-recursion off.
Developers: this version has a new pdns-2.1 development kit, available on http://downloads.powerdns.com/releases/dev. See also Appendix C.
![]() | Most users will run a static version of PDNS which has no dependencies on external libraries. However, some may need to run the dynamic version. This warning applies to these users. To run the dynamic version of PDNS, which is needed for backend drivers which are only available in source form, gcc 3.0 is required. RedHat 7.2 comes with gcc 3.0 as an optional component, RedHat 7.3 does not. However, the RedHat 7.2 Update gcc rpms install just fine on RedHat 7.3. For Debian, we suggest running 'woody' and installing the g++-3.0 package. We expect to release a FreeBSD dynamic version shortly. |
Bugs fixed:
RPM releases sometimes overwrote previous configuration files. Thanks to Jorn Ekkelenkamp of Hubris/ISP Services for reporting this.
TCP recursion sent out overly large responses due to a byteorder mistake, confusing some clients. Thanks to the capable engineers of NetNation for bringing this to our attention.
TCP recursion in combination with a recursing backend on a non-standard port did not work, leading to a non-functioning TCP listener. Thanks to the capable engineers of NetNation for bringing this to our attention.
Unexpected behaviour:
Wildcard URL records where not implemented because they are a performance penalty. To turn these on, enable wildcard-url in the configuration.
Unlike other nameservers, local data did not override the internet for recursing queries. This has mostly been brought into conformance with user expectations. If a recursive question can be answered entirely from local data, it is. To restore old behaviour, disable lazy-recursion. Also see Chapter 11.
Features:
Oracle support has been tuned, leading to the first public release of the Oracle backend. Zone2sql now outputs better SQL and the backend is now fully documented. Furthermore, the queries are compatible with the PowerDNS XML-RPC product, allowing PowerDNS express to run off Oracle. See Section A.6.
Zone2sql now accepts --transactions to wrap zones in a transaction for PostgreSQL and Oracle output. This is a major speedup and also makes for better isolation of inserts. See Section 10.1.
pdns_control now has the ability to purge the PowerDNS cache or parts of it. This enables operators to raise the TTL of the Packet Cache to huge values and only to invalidate the cache when changes are made. See also Chapter 9 and Section B.1.1.
Maintenance release, fixing three small issues.
Developers: this version is compatible with 1.99.11 backends.
PowerDNS ignored the logging-facility setting unless it was specified on the commandline. Thanks to Karl Obermayer from WebMachine Technologies for noticing this.
Zone2sql neglected to preserve 'slaveness' of domains when converting to the slave capable PostgreSQL backend. Thanks to Mike Benoit of NetNation for reporting this. Zone2sql now has a --slave option.
SOA Hostmaster addresses with dots in them before the @-sign were mis-encoded on the wire.
Two bugfixes, one stability/security related. No new features.
Developers: this version is compatible with 1.99.11 backends.
Bugfixes:
zone2sql refused to work under some circumstances, taking 100% cpu and not functioning. Thanks to Andrew Clark and Mike Benoit for reporting this.
Fixed a stability issue where malformed packets could force PDNS to reload. Present in all earlier 2.0 versions.
Mostly bugfixes, no really new features.
Developers: this version is compatible with 1.99.11 backends.
Bugs fixed:
chroot() works again - 2.0rc1 silently refused to chroot. Thanks to Hub Dohmen for noticing this.
setuid() and setgid() security features were silently not being performed in 2.0rc1. Thanks to Hub Dohmen for noticing this.
MX preferences over 255 now work as intended. Thanks to Jeff Crowe for noticing this.
IPv6 clients can now also benefit from the recursing backend feature. Thanks to Andy Furnell for proving beyond any doubt that this did not work.
Extremely bogus code removed from DNS notification reception code - please test! Thanks to Jakub Jermar for working with us in figuring out just how broken this was.
AXFR code improved to handle more of the myriad different zonetransfer dialects available. Specifically, interoperability with Bind 4 was improved, as well as Bind 8 in 'strict rfc conformance' mode. Thanks again for Jakub Jermar for running many tests for us. If your transfers failed with 'Unknown type 14!!' or words to that effect, this was it.
Features:
Win32 version now has a zone2sql tool.
Win32 version now has support for specifying how urgent messages should be before they go to the NT event log.
Remaining issues:
One persistent report of the default 'chroot=./' configuration not working.
One report of disable-axfr and allow-axfr-ips not working as intended.
Support for relative paths in zones and in Bind configuration is not bug-for-bug compatible with bind yet.
The MacOS X release! A very experimental OS X 10.2 build has been added. Furthermore, the Windows version is now in line with Unix with respect to capabilities. The ODBC backend now has the code to function as both a master and a slave.
Developers: this version is compatible with 1.99.11 backends.
Implemented native packet response parsing code, allowing Windows to perform AXFR and NS and SOA queries.
This is the first version for which we have added support for Darwin 6.0, which is part of the forthcoming Mac OS X 10.2. Please note that although this version is marked RC1, that we have not done extensive testing yet. Consider this a technology preview.
The Darwin version has been developed on Mac OS X 10.2 (6C35). Other versions may or may not work.
Currently only the random, bind, mysql and pdns backends are included.
The menu based installer script does not work, you will have to edit pathconfig by hand as outlined in chapter 2.
On Mac OS X Client, PDNS will fail to start because a system service is already bound to port 53.
This version is distributed as a compressed tar file. You should follow the generic UNIX installation instructions.
Bugs fixed:
Zone2sql PostgreSQL mode neglected to lowercase $ORIGIN. Thanks to Maikel Verheijen of Ladot for spotting this.
Zone2sql PostgreSQL mode neglected to remove a trailing dot from $ORIGIN if present. Thanks to Thanks to Maikel Verheijen of Ladot for spotting this.
Zonefile parser was not compatible with bind when $INCLUDING non-absolute filenames. Thanks to Jeff Miller for working out how this should work.
Bind configuration parser was not compatible with bind when including non-absolute filenames. Thanks to Jeff Miller for working out how this should work.
Documentation incorrectly listed the Bind backend as 'slave capable'. This is not yet true, now labeled 'experimental'.
Windows changes. We are indebted to Dimitry Andric who educated us in the ways of distributing Windows software.
pdns.conf is now read if available.
Console version responds to ^c now.
Default pdns.conf added to distribution
Uninstaller missed several files, leaving remnants behind
DLLs are now installed locally, with the pdns executable.
pdns_control is now also available on Windows
ODBC backend can now act as master and slave. Experimental.
The example zone missed indexes and had other faults.
A runtime DLL that is present on most windows systems (but not all!) was missing.
The Windows release! See Chapter 3. Beware, windows support is still very fresh and untested. Feedback is very welcome.
Developers: this version is compatible with 1.99.11 backends.
Windows 2000 codebase merge completed. This resulted in quite some changes on the Unix end of things, so this may impact reliability
ODBC backend added for Windows. See Section A.10.
IBM DB2 Universal Database backend available for Linux. See Section A.8.
Zone2sql now understands $INCLUDE. Thanks to Amaze Internet for nagging about this
The SOA Mininum TTL now has a configurable default (soa-minimum-ttl)value to placate the DENIC requirements.
Added a limit on the simultaneous numbers of TCP connections to accept (max-tcp-connections). Defaults to 10.
Bugs fixed:
When operating in virtual hosting mode (See Chapter 8), the additional init.d scripts would not function correctly and interface with other pdns instances.
PDNS neglected to conserve case on answers. So a query for WwW.PoWeRdNs.CoM would get an answer listing the address of www.powerdns.com. While this did not confuse resolvers, it is better to conserve case. This has semantical concequences for all backends, which the documentation now spells out.
PostgreSQL backend was case sensitive and returned only answers in case an exact match was found. The Generic PostgreSQL backend is now officially all lower case and zone2sql in PostgreSQL mode enforces this. Documentation has been been updated to reflect the case change. Thanks to Maikel Verheijen of Ladot for spotting this!
Documentation bug - postgresql create/index statements created a duplicate index. If you've previously copy pasted the commands and not noticed the error, execute CREATE INDEX rec_name_index ON records(name) to remedy. Thanks to Jeff Miller for reporting this. This also lead to depressingly slow 'ANY' lookups for those of you doing benchmarks.
Features:
pdns_control (see Section B.1.1) now opens the local end of its socket in /tmp instead of next to the remote socket (by default /var/run). This eases the way for allowing non-root access to pdns_control. When running chrooted (see Chapter 7), the local socket again moves back to /var/run.
pdns_control now has a 'version' command. See Section B.1.1.
This release is important because it is the first release which is accompanied by an Open Source Backend Development Kit, allowing external developers to write backends for PDNS. Furthermore, a few bugs have been fixed:
Lines with only whitespace in zone files confused PDNS (thanks Henk Wevers)
PDNS did not properly parse TTLs with symbolic sufixes in zone files, ie 2H instead of 7200 (thanks Henk Wevers)
IMPORTANT: there has been a tiny license change involving free public webbased dns hosting, check out the changes before deploying!
PDNS is now feature complete, or very nearly so. Besides adding features, a lot of 'fleshing out' work is done now. There is an important performance bug fix which may have lead to disappointing benchmarks - so if you saw any of that, please try either this version or 1.99.8 which also does not have the bug.
This version has been very stable for us on multiple hosts, as was 1.99.9.
PostgreSQL users should be aware that while 1.99.10 works with the schema as presented in earlier versions, advanced features such as master or slave support will not work unless you create the new 'domains' table as well.
Bugs fixed:
Wildcard AAAA queries sometimes received an NXDOMAIN error where they should have gotten an empty NO ERROR. Thanks to Jeroen Massar for spotting this on the .TK TLD!
Do not disable the packetcache for 'recursion desired' packets unless a recursor was configured. Thanks to Greg Schueler for noticing this.
A failing backend would not be reinstated. Thanks to 'Webspider' for discovering this problem with PostgreSQL connections that die after prolonged inactivity.
Fixed loads of IPv6 transport problems. Thanks to Marco Davids and others for testing. Considered ready for production now.
Zone2sql printed a debugging statement on range $GENERATE commands. Thanks to Rene van Valkenburg for spotting this.
Features:
PDNS can now act as a master, sending out notifications in case of changes and allowing slaves to AXFR. Big rewording of replication support, domains are now either 'native', 'master' or 'slave'. See Chapter 13 for lots of details.
Zone2sql in PostgreSQL mode now populates the 'domains' table for easy master, slave or native replication support.
Ability to disable those annoying Windows DNS Dynamic Update messages from appearing in the log. See log-failed-updates
in Chapter 15.
Ability to run on IPv6 transport only
Logging can now happen under a 'facility' so all PDNS messages appear in their own file. See Section 6.3.
Different OS releases of PDNS now get different install path defaults. Thanks to Mark Lastdrager for nagging about this and to Nero Imhard and Frederique Rijsdijk for suggesting saner defaults.
Infrastructure for 'also-notify' statements added.
This is again a feature and an infrastructure release. We are nearly feature complete and will soon start work on the backends to make sure that they are all master, slave and 'superslave' capable.
Bugs fixed:
PDNS sometimes sent out duplicate replies for packets passed to the recursing backend. Mostly a problem on SMP systems. Thanks to Mike Benoit for noticing this.
Out-of-bailiwick CNAMES (ie, a CNAME to a domain not in PDNS) caused a 'ServFail' packet in 1.99.8, indicating failure, leading to hosts not resolving. Thanks to Martin Gillstrom for noticing this.
Zone2sql balked at zones editted under operating sytems terminating files with ^Z (Windows). Thanks Brian Willcott for reporting this.
PostgreSQL backend logged the password used to connect. Now only does so in case of failure to connect. Thanks to 'Webspider' for noticing this.
Debian unstable distribution wrongly depended on home compiled PostgreSQL libraries. Thanks to Konrad Wojas for noticing this.
Features:
When operating as a slave, AAAA records are now supported in the zone. They were already supported in master zones.
IPv6 transport support - PDNS can now listen on an IPv6 socket using the local-ipv6 setting.
Very silly randombackend added which appears in the documentation as a sample backend. See Appendix C.
When transferring a slave zone from a master, out of zone data is now rejected. Malicious operators might try to insert bad records otherwise.
'Supermaster' support for automatic provisioning from masters. See Section 13.2.1.
Recursing backend can now live on a non-standard (!=53) port. See Chapter 11.
Slave zone retrieval is now queued instead of immediate, which scales better and is more resilient to temporary failures.
max-queue-length parameter. If this many packets are queued for database attention, consider the situation hopeless and respawn.
Internal:
SOA records are now 'special' and each backend can optionally generate them in special ways. PostgreSQL backend does so when operating as a slave.
Writing backends is now a lot easier. See Appendix C.
Added Bindbackend to internal regression tests, confirming that it is compliant.
A lot of infrastructure work gearing up to 2.0. Some stability bugs fixed and a lot of new features.
Bugs fixed:
Bindbackend was overly complex and crashed on some systems on startup. Simplified launch code.
SOA fields were not always properly filled in, causing default values to go out on the wire
Obscure bug triggered by malicious packets (we know who you are) in SOA finding code fixed.
Magic serial number calculation contained a double free leading to instability.
Standards violation, questions for domains for which PDNS was unauthoritative now get a SERVFAIL answer. Thanks to the IETF Namedroppers list for helping out with this.
Slowly launching backends were being relaunched at a great rate when queries were coming in while launching backends.
MySQL-on-unix-domain-socket on SMP systems was overwhelmed by the quick connection rate on launch, inserted a small 50ms delay.
Some SMP problems appear to be compiler related. Shifted to GCC 3.0.4 for Linux.
Ran ispell on documentation.
Feature enhancements:
Recursing backend. See Chapter 11. Allows recursive and authoritative DNS on the same IP address.
NAPTR support, which is especially useful for the ENUM/E.164 community.
Zone transfers can now be allowed per netmask instead of only per IP address.
Preliminary support for slave operation included. Only for the adventurous right now! See Section 13.2
All record types now documented, see Chapter 17.
Wildcard CNAMES do not work as they do with bind.
Recursion sometimes sends out duplicate packets (fixed in 1.99.9 snapshots)
Some stability issues which are caught by the guardian
Features present in this document, but disabled or withheld from the current release:
gmysqlbackend, oraclebackend
Named.conf parsing got a lot of work and many more bind configurations can now be parsed. Furthermore, error reporting was improved. Stability is looking good.
Bugs fixed:
Bind parser got confused by filenames with underscores and colons.
Bind parser got confused by spaces in quoted names
FreeBSD version now stops and starts when instructed to do so.
Wildcards were off by default, which violates standards. Now on by default.
--oracle was broken in zone2sql
Feature enhancements:
Line number counting goes on as it should when including files in named.conf
Added --no-config to enable users to start the pdns daemon without parsing the configuration file.
zone2sql now has --bare for unformatted output which can be used to generate insert statements for different database layouts
zone2sql now has --gpgsql, which is an alias for --mysql, to output in a format useful for the default Generic PgSQL backend
zone2sql is now documented.
Wildcard CNAMES do not work as they do with bind.
Features present in this document, but disabled or withheld from the current release:
gmysqlbackend, oraclebackend
This version is now running on dns-eu1.powerdns.net and working very well for us. But please remain cautious before deploying!
Bugs fixed:
Webserver neglected to show log messages
TCP question/answer miscounted multiple questions over one socket. Fixed misnaming of counter
Packetcache now detects clock skew and times out entries
named.conf parser now reports errors with line number and offending token
Filenames in named.conf can now contain :
Feature enhancements:
The webserver now by default does not print out configuration statements, which might contain database backends. Use webserver-print-arguments to restore the old behaviour.
Generic PostgreSQL backend is now included. Still rather beta.
FreeBSD version does not stop when requested to do so.
Wildcard CNAMES do not work as they do with bind.
Features present in this document, but disabled or withheld from the current release:
gmysqlbackend, oraclebackend
The main focus of this release is stability and TCP improvements. This is the first release PowerDNS-the-company actually considers for running on its production servers!
Major bugs fixed:
Zone2sql received a floating point division by zero error on named.confs with less than 100 domains.
Huffman encoder failed without specific error on illegal characters in a domain
Fixed huge memory leaks in TCP code.
Removed further file descriptor leaks in guardian respawning code
Pipebackend was too chatty.
pdns_server neglected to close fds 0, 1 & 2 when daemonizing
Feature enhancements:
bindbackend can be instructed not to check the ctime of a zone by specifying bind-check-interval=0, which is also the new default.
pdns_server --list-modules lists all available modules.
Performance enhancements:
TCP code now only creates a new database connection for AXFR.
TCP connections timeout rather quickly now, leading to less load on the server.
FreeBSD version does not stop when requested to do so.
Wildcard CNAMES do not work as they do with bind.
Features present in this document, but disabled or withheld from the current release:
gmysqlbackend, oraclebackend, gpgsqlbackend
A lot of new named.confs can now be parsed, zone2sql & bindbackend have gained features and stability.
Major bugs fixed:
Label compression was not always enabled, leading to large reply packets sometimes.
Database errors on TCP server lead to a nameserver reload by the guardian.
MySQL backend neglected to close its connection properly.
BindParser miss parsed some IP addresses and netmasks.
Truncated answers were also truncated on the packetcache, leading to truncated TCP answers.
Feature enhancements:
Zone2sql and the bindbackend now understand the Bind $GENERATE{} syntax.
Zone2sql can optionally gloss over non-existing zones with --on-error-resume-next.
Zone2sql and the bindbackend now properly expand @ also on the right hand side of records.
Zone2sql now sets a default TTL.
DNS UPDATEs and NOTIFYs are now logged properly and sent the right responses.
Performance enhancements:
'Fancy records' are no longer queried for on ANY queries - this is a big speedup.
FreeBSD version does not stop when requested to do so.
Zone2sql refuses named.confs with less than 100 domains.
Wildcard CNAMES do not work as they do with bind.
Features present in this document, but disabled or withheld from the current release:
gmysqlbackend, oraclebackend, gpgsqlbackend
The big news in this release is the BindBackend which is now capable of parsing many more named.conf Bind configurations. Furthermore, PDNS has successfully parsed very large named.confs with large numbers of small domains, as well as small numbers of large domains (TLD).
Zone transfers are now also much improved.
Major bugs fixed:
zone2sql leaked file descriptors on each domain, used wrong Bison recursion leading to parser stack overflows. This limited the amount of domains that could be parsed to 1024.
zone2sql can now read all known zonefiles, with the exception of those containing $GENERATE
Guardian relaunching a child lost two file descriptors
Don't die on a connection reset by peer during zone transfer.
Webserver does not crash anymore on ringbuffer resize
Feature enhancements:
AXFR can now be disabled, and re-enabled per IP address
--help accepts a parameter, will then show only help items with that prefix.
zone2sql now accepts a --zone-name parameter
BindBackend maturing - 9500 zones parsed in 3.5 seconds. No longer case sensitive.
Performance enhancements:
Implemented RFC-breaking AXFR format (which is the industry standard). Zone transfers now zoom along at wirespeed (many megabits/s).
FreeBSD version does not stop when requested to do so.
BindBackend cannot parse zones with $GENERATE statements.
Features present in this document, but disabled or withheld from the current release:
gmysqlbackend, oraclebackend, gpgsqlbackend
Major bugs fixed:
Database backend reload does not hang the daemon anymore
Buffer overrun in local socket address initialisation may have caused binding problems
setuid changed the uid to the gid of the selected user
zone2sql doesn't coredump on invocation anymore. Fixed lots of small issues.
Don't parse configuration file when creating configuration file. This was a problem with reinstalling.
removed a lot of unnecessary gettimeofday calls
removed needless select(2) call in case of listening on only one address
removed 3 useless syscalls in the fast path
Usability improvements:
Fixed error checking in init.d script (show, mrtg)
Added 'uptime' to the mrtg output
removed further GNUisms from installer and init.d scripts for use on FreeBSD
Debian package and apt repository, thanks to Wichert Akkerman.
FreeBSD /usr/ports, thanks to Peter van Dijk (in progress).
Stability may be an issue as well as performance. This version has a tendency to log a bit too much which slows the nameserver down a lot.
Decreasing a ringbuffer on the website is a sure way to crash the daemon. Zone2sql, while improved, still has problems with a zone in the following format:
name IN A 1.2.3.4 IN A 1.2.3.5To fix, add 'name' to the second line.
Zone2sql does not close filedescriptors.
FreeBSD version does not stop when requested via the init.d script.
Features present in this document, but disabled or withheld from the current release:
gmysqlbackend, oraclebackend, gpgsqlbackend
fully functioning bindbackend - will try to parse named.conf, but probably fail
This is the first public release of what is going to become PDNS 2.0. As such, it is not of production quality. Even PowerDNS-the-company does not run this yet.
Stability may be an issue as well as performance. This version has a tendency to log a bit too much which slows the nameserver down a lot.
Decreasing a ringbuffer on the website is a sure way to crash the daemon. Zone2sql is very buggy.
Features present in this document, but disabled or withheld from the current release:
gmysqlbackend, oraclebackend, gpgsqlbackend
fully functioning bindbackend - will not parse configuration files