Krill

Krill is a Resource Public Key Infrastructure (RPKI) daemon, featuring a Certificate Authority (CA) and publication server, written in Rust. If you have any feedback, we would love to hear from you. Don’t hesitate to create an issue on Github or post a message on our RPKI mailing list or Discord server.

For more information please refer to the documentation.

Public Krill Testbed Service

In case you want to try out Krill, or test new Krill releases, then you may want to use our public Krill testbed service. You can read more about this service in this blog post.

Changelog

0.14.4 'A Flock of Krill'

This release fixes the following issues:

• Krill should not freeze if lockfiles were not deleted properly #1171 (since Krill 0.14.0)
• Don't warn about yanked dependencies when installing Krill via Cargo #1173

0.14.3 'Temp'

This release fixes a number of issues found in 0.14.0 through 0.14.2:

• Use rpki-rs 0.18.0 to support builds on more platforms #1166
• Fix aspa migration issues #1163
• Depend on kvx 0.9.2 to ensure temp files are used properly #1160

Most importantly, Krill will now use temp files for all data that it stores to avoid issues with half-written files in case the disk is full, or the server is rebooted in the middle of writing. This issue was introduced in release 0.14.0, and we recommend that all users upgrade to this version to avoid issues.

This release also includes:

• Updated German UI translations krill-ui/#51

0.14.2 'Extra, Extra, Extra!'

This release fixes an additional corner case in the migration code that affects certain installations that archived 'surplus' commands (issue: #1147). There is no need to upgrade to this version if you already upgraded to 0.14.0 or 0.14.1.

0.14.1 'Extra, Extra!'

This release fixes a bug in the migration code that affects certain installations that archived 'surplus' data (issue: #1147). There is no need to upgrade to this version if you already upgraded to 0.14.0.

0.14.0 'ASPA'

This release adds support for the updated ASPA v1 profile (issue #1080). Any existing ASPA objects will be re-issued automatically.

In addition, the following small features and fixes were done:

• Show delete ROA button when no BGP preview is available #1139
• Add traditional and simplified Chinese translations #1075
• Let the testbed automatically renew the TA manifest and CRL #1095 (see below)
• Show the delete icon for AS0 ROA when there is another existing announcement #1109

The main effort in this release was spent on less user-visible improvements in the way that Krill stores its data. This will help to improve robustness today, and it paves the way for introducing support for Krill clustering using a database back-end in a future release.

For now, these issues have been done:

• Improve transactionality of changes (e.g. #1076-1078, #1085, #1108, #1090)
• Remove no longer needed 'always_recover_data' function #1086
• Improve upgrade failed error: tell users to downgrade #1042
• Crash Krill if the task scheduler encounters a fatal error. #1132

You can find the full list of issues here: https://github.com/NLnetLabs/krill/projects/25

Finally, regarding issue #1095. If you were running 0.13.1 as a testbed, you may have symlinked the "signer" directory to "ta_signer" to support a manual workaround for re-signing the trust anchor CRL and manifest. If you did, you may need to delete any surplus files and directories under "data/ta_signer" other than the directory called "ta".

0.13.1 'Scrollbars!'

The Krill UI includes a CA selection dropdown in case you have multiple CAs. This dropdown used to have a scrollbar, which accidentally got lost in the UI overhaul we did in version 0.13.0. This is now fixed (#1071)

0.13.0 'DRY'

Summary

This release contains an important fix for an issue affecting v0.12.x Publication Servers (see PR #1023). It is recommended that affected installations are upgraded as soon as possible.

The user interface was completely re-implemented in this release resulting in a smaller browser footprint. Functionality is mostly unchanged, except that users can now have an optional comment with each of their ROA configurations. These comments are not part of published ROA objects - they are meant for local bookkeeping only.

ASPA objects are now supported through the CLI by default. We hope to add UI support later this year.

Krill can now be used as a full RPKI Trust Anchor, using a detached (possibly offline) signer for Trust Anchor key operations.

Publication Server

Krill 0.12.x Publication servers suffer from an issue where multiple entries for the same URI, but with different hashes can appear in a single RRDP snapshot.

This problem was solved by removing published objects data duplication in the Krill architecture and ensuring that the URI rather than an object's hash is used as its primary key internally. More information can be found in pull request #1023.

We recommend that existing 0.12.x Publication Server installations are upgraded to this version.

Updated User Interface

A lot of changes were introduced in this release. For most users the following improvements will be most visible and relevant:

• Updated UI to new and smaller code base (#995)
• Allow ROA comments in UI (#995)

The new krill-ui project has its own repository where issues can be tracked: https://github.com/NLnetLabs/krill-ui

ASPA Support

ASPA support is now enabled in the CLI (#1031). We hope to add UI support later this year.

We added a number of new restrictions

• Krill MUST NOT create only a single AFI ASPA (#1063)
• ASPA object MUST NOT allow the customer AS in the provider AS list (#1058)

You can read more about ASPA support here: https://krill.docs.nlnetlabs.nl/en/0.13.0/manage-aspas.html

API Changes

We removed the repository next update time from the stats and metrics output. It was inaccurate (usually 8 hours off), and not very informative. More useful metrics are still provided: last exchange and last successful exchange. If these times differ, then there is an issue that may need attention.

Krill as a Trust Anchor

A lot of work has been done to support using Krill as a Trust Anchor. If you are not an RIR, then you will not need to run your own RPKI TA for normal RPKI operations. That said, some users may want to operate their own TA outside of the TAs provided by the RIRs for testing, study or research reasons. Or perhaps even to manage private use address space.

You can read more about this here: https://krill.docs.nlnetlabs.nl/en/0.13.0-rc1/trust-anchor.html

Implemented issues:

• Support offline TA (#976)
• Support initialising offline TA with existing key (#979)
• Bulk import/configure CAs with ROAs (#968, #969)
• Support migration of existing TAs (#978)
• Use new TA for embedded (test) TA (#977)

Other Changes

Publication Server Improvements:

• Remove published object data duplication (#1023)
• Delete repository files by URI (#991)

Miscellaneous improvements and fixes:

• Log for which child / parent / publisher CMS validation failed (#1027)
• Permit setting CKA_PRIVATE to CK_FALSE on PKCS#11 RSA public keys (#1019)
• Ensure that the CSR uses a trailing slash for id-ad-caRepository (#1030)
• Accept id-cert with path len constraints (#966)
• Publication Server should check uri, not hash, in publish elements (#981)

The overview of all issues for this release can be found here: https://github.com/NLnetLabs/krill/projects/24

0.12.3 'Sakura'

This release contains a feature that enables Publication Server operators to remove unwanted, surplus, files from their repository. This feature was cherry picked from the upcoming major release branch so that Publication Server operators can use this without delay.

Note that if you do not use Krill to operate a Publication Server, then there is no need to upgrade to this version now.

For more details see: https://github.com/NLnetLabs/krill/pull/1022

0.12.2 'Dijkstra'

This release fixes a locking issue that can affect a Krill Publication Server with a large number of concurrent publishers. See PR #1007.

If you only use Krill as an RPKI Certificate Authority and publish elsewhere, e.g. in an RPKI Publication Server provided by your RIR or NIR, then there is no need to update to this release.

0.12.1 'Safety Belts'

This release introduces two fixes for the Krill Publication Server. If you only use Krill as an RPKI Certificate Authority and publish elsewhere, e.g. in an RPKI Publication Server provided by your RIR or NIR, then there is no need to update to this release.

Firstly, this release fixes CVE-2023-0158: https://nlnetlabs.nl/downloads/krill/CVE-2023-0158.txt

This CVE describes an exposure where remote attackers could cause Krill to crash if it is used as an RPKI Publication Server and if its "/rrdp" endpoint is accessible over the public internet.

Note that servers are not affected if the advice in our documentation was followed and a separate web server is used to serve the RRDP data:

https://krill.docs.nlnetlabs.nl/en/stable/publication-server.html#synchronise-repository-data

Secondly, locking was added in this release to ensure that updates to the repository content are always applied sequentially. This fixes a concurrency issue introduced in Krill 0.12.0 that could result in rejecting an update from a publishing CA. In such cases the affected update would not be visible for RPKI validators, until a later publication attempt would be successful.

We advise that users upgrade to this version of Krill if they use it as their RPKI Publication Server. We also continue to recommend that a separate web server is used for serving the RRDP data.

0.12.0 'Crickets'

This release vastly reduces the CPU usage by Publication Servers for big RPKI repositories.

In addition to this we added a small feature, and fixed an interop issue:

• Listen on IPv4+IPv6 #955
• Fix rfc6492 interop (AKI format) #948

Upgrade instructions this release are here: https://krill.docs.nlnetlabs.nl/en/stable/upgrade.html#v0-12-0

The overview of all issues for this release can be found here: https://github.com/NLnetLabs/krill/projects/23

Full documentation can be found here: https://krill.docs.nlnetlabs.nl/

0.11.0 'What about that ROA?'

In this release we introduce two features in the Krill API and CLI:

• Support optional comment for each ROA configuration #863
• Show ROA object(s) for each ROA configuration #864

This is not yet supported in the UI, but will be in the near future as the current UI will get a make-over soon.

Other than this we included a few minor issues and fixes:

• Query initialisation parameters for Krill pubserver (rrdp/rsync URI) #835
• Tasks for removed CAs should not result in errors #906
• Disallow negative numbers in config #808

Documentation can be found here: https://krill.docs.nlnetlabs.nl/en/stable/

And here: https://krill.docs.nlnetlabs.nl/en/stable/upgrade.html#v0-11-0

0.10.3 'Down Under'

This release fixes an interoperability issue with the APNIC CA system which didn't occur in the public test environment. See issue #933.

0.10.2 'All Types'

This release fixes an issue where Krill 0.10.0 and 0.10.1 parent CAs would issue an RFC 6492 Resource Class List Response with a missing, rather than empty, attribute for resource types that a child CA had no entitlements for. See issue #925.

Note this issue did not impact CAs that have no delegated child CAs. Futhermore, it would result in the child CA rejecting the parent response, log an error, and try synchronising again, but it would not result in any changes to the CA certificate issued by the parent to the child.

Furthermore, this release fixes an issue where the data for a Publication Server (if configured) would not be migrated if the previous Krill version was 0.9.0. See issue #928.

0.10.1 'Slash'

Krill 0.10.0, or rather rpki-rs 0.15.4 became quite strict in its validation of the RFC 8183 XML files used when setting up a CA. As a result Krill 0.10.0 rejected the XML generated by earlier versions because the trailing slash in the XML namespace was missing.

Because the namespace is not critical in this context, this new Krill release will no longer reject the XML files because of this missing trailing slash.

0.10.0 'Hush'

In this release we introduce the following major features:

• BGPSec Router Certificate Signing
• Support the use of Hardware Security Modules (HSMs) for key operations

The documentation has more information:

Besides these major features we added a number of small improvements and bugfixes:

• CRL revocation dates in the future #788
• Prevent that two krill instances modify the same data #829
• Let user force RRDP session reset on restore #828
• Various code improvements aimed at maintainability
• Using a jitter of 0 results in a panic #859
• Security fixes in KMIP dependencies #860 (HSM support)
• Add SSLKEYLOGFILE support #615
• Allow explicit disabling of HTTPS #913

The full list of changes can be found here: https://github.com/NLnetLabs/krill/projects/19

0.9.6 'Newer ROAs Please'

This release fixes an issue introduced in 0.9.5 where the background job to automatically renew ROAs was not added to Krill's task queue on startup. Thanks to Alberto Leiva for finding this issue!

All users who upgraded to 0.9.5 are advised to upgrade to this version as soon as possible. Not doing so can lead to ROAs expiring and becoming invalid. If you did not upgrade to 0.9.5 you are not affected by this issue.

This release contains no other changes.

0.9.5 'Have You considered these Upgrades?'

This release is primarily intended to improve support for migrations of pre-0.9.0 installations. The upgrade code has been separated more cleanly into a step where the new 0.9.0 data structures are prepared in a new directory first, and a second step where this new data is made active and the old data is archived. Earlier versions of krill were performing data migrations in-place.

If you simply upgrade krill and restart it, then it will automatically execute both steps. If the preparation step should fail, then the original data remains unchanged. You can then downgrade back to your previous krill version. This is in itself is an improvement over 0.9.4 and earlier, because for those versions you would have to make a back-up of your data first, and restore it in order to revert your upgrade.

Furthermore, we have now added a new command line tool called 'krillup', which can be installed and upgraded separately to krill itself. This new tool can be used to execute the krill migration preparation step only. Meaning, you can install this tool on your server and do all the preparations, and only then upgrade krill.

This has the following advantages:

• The downtime for data migrations is reduced for servers with lots of data
• If the preparation fails, there is no need to revert a krill update

In addition to this we have also made some changes to the CA parent refresh logic. Krill CAs were checking their entitlements with their parents every 10 minutes, and this causes too much load on parent CAs with many children. There should be no need to check this often. CAs will now check every 24 to 36 hours, using a random spread. This will decrease the load on parent CAs significantly.

Note that you can always force a 'parent refresh' sooner through the UI or command line (krillc bulk refresh). You may want to use this if your parent informs you through other channels that your resources have changed - e.g. you were allocated a new prefix.

Secondly, because the next synchronisation time is now difficult to predict in the code that reports the parent status - it is now no longer shown in the UI/API. We may add this back in a future release. See issue #807.

You can read more about this upgrade process here: https://krill.docs.nlnetlabs.nl/en/latest/upgrade.html

In addition to this we added a few other quick fixes in this release:

• Make RRDP session reset manual option #793
• Improve http connection error reporting #776
• Fix deserialization bug for CAs with children #774
• Connect to local parent directly #791
• Do not sign/validate RFC6492 messages to/from local parent #797
• Use per CA locking for CA statuses #795
• Decrease CA update frequency and use jitter to spread load #802
• Accept missing tag in RFC8181 Error Response #809
• Improve efficiency of connection status tracking #811
• Do not resync CAs with repo on startup if there are too many #818

The full list of changes can be found here: https://github.com/NLnetLabs/krill/projects/20

0.9.4 'One shall be the number thou shalt count from'

This release includes the following:

• RRDP serial should start from 1, not 0 (#741)
• Allow configuring RFC6492/8181 client timeouts (#743)

The first addresses a non-critical bug found when running Krill as a Publication Server. The second addresses an issue seen in Krill 0.7.3 running with 100s of CAs in a single Krill instance - timeouts have not been seen in Krill 0.9.x - but it does not hurt to give operators control over this configuration.

If you are using Krill for RPKI CA functions only, and you have already upgraded to version 0.9.3 then there is no immediate need to upgrade to this version. If you are running a version from before 0.9.3, then you are still advised to upgrade to this version for the reasons list under version 0.9.3.

0.9.3 'The Thundering Herd'

This release adds the following features and fixes:

• Prevent a thundering herd of hosted CAs publishing at the same time (#692)
• Re-issue ROAs to ensure that short EE subject names are used (#700)
• Handle rate limits when updating parents (#680)
• Support experimental ASPA objects through CLI (#685)

Note that ASPA objects are not intended for use in production environments just yet. We have added experimental support for this to support the development of the ASPA standards in the IETF. Information on how to use Krill to manage ASPA objects can be found here: https://krill.docs.nlnetlabs.nl/en/prototype-aspa-support/manage-aspas.html

The full list of changes can be found here: https://github.com/NLnetLabs/krill/projects/18

0.9.2 'Motive and Opportunity'

This release includes two features aimed at users who run a Krill CA to maintain ROAs:

• Warn about ROA configurations for resources no longer held #602
• Re-enable migration of CA content to a new Publication Server #480

In addition to this we have added a lot of smaller improvements:

• Synchronize the manifest EE lifetime and next update time #589
• Improve error reporting on I/O errors #587
• Add rsync URI to testbed TAL #624
• Improve status reporting and monitoring #651, #650, #648

The following features were added to support users who operate Krill as a parent CA, or Publication Server:

• Optionally suspend inactive child CAs using krill 0.9.2 and up #670
• Perform RRDP session reset on restart #533
• Use unguessable URIs for RRDP deltas and snapshots #515

The updated documentation for this release can be found here: https://krill.docs.nlnetlabs.nl/en/0.9.2/index.html

The full list of changes can be found here: https://github.com/NLnetLabs/krill/projects/16

0.9.1 'All for One'

This release fixes an issue where the Publication Server would lock up (#606). Users who do not use Krill to operate their own Publication Server do not need to upgrade to this release.

This locking issue was cause by slow deserialisation of the repository content. It primarily affected large repositories because more content makes this process slower, and having more publishers who publish regularly means it is triggered more frequently.

0.9.0 'One for All'

This is the first major release of Krill in a while.

While basic ROA management is unchanged, there were many changes under the hood:

• Multi-user support in the User Interface (local users or OpenID Connect)
• Reduce disk space usage and growth over time
• API and naming consistency (in preparation for 1.0 in future)
• Publication Server improvements (to whom it may concern)
• Many small improvements and minor bug fixes

For a full list of issues that were included in this release see: https://github.com/NLnetLabs/krill/projects/4

Updated documentation is available here: https://krill.docs.nlnetlabs.nl/en/stable/index.html

With multi-user support you can now give people in your organization individual access rights to your CA - and they no longer need to share a password. If you have an OpenID Connect provider then you can integrate Krill with it. Read more here: https://krill.docs.nlnetlabs.nl/en/stable/multi-user.html

Krill versions before 0.9.0 keep a lot of data around that is not strictly needed. This can clog up your system and it makes the Krill history difficult to parse. History can seen using krillc history. We will include support for inspecting history in the UI soon.

There were some API and CLI changes introduced in this release. Over time things had become a bit inconsistent and we felt we needed to fix that before we can consider going for the Krill 1.0 release. If you are using automation then these changes may break your current integrations. Please have a look at the following page to see if and how this affects you: https://krill.docs.nlnetlabs.nl/en/stable/upgrade.html

Note that your Krill data store will be upgraded automatically if you upgrade to this release. This upgrade can take some time, up to around 30 minutes dependent on the amount of history which accumulated over time and the speed of your system. During the migration you will not be able to update your ROAs, but your existing ROAs will remain available to RPKI validators. I.e. there is no downtime expected with regards to RPKI validation.

We have tested this on various (big) Krill instances running CAs as well as Publication Servers. Still, we recommend that you make a backup of your data store before upgrading. In case the upgrade should unexpectedly fail for you, please restore your old data, run the previous binary, and contact us so that we can make a fix. Alternatively, copy your data except for the keys directory to a test system and then use the new Krill binary there with the following env variable set so you can test the data migration:

KRILL_UPGRADE_ONLY=1

Finally, note that you need to run at least Krill 0.6.0 in order to upgrade. If you run an older version you will need to upgrade to version 0.8.2 first.

0.8.2 'Can't touch this'

As it turned out the previous release (0.8.1) still insisted on cleaning up 'redundant ROAs' when migrating to that version. This clean-up would not cause any issues with regards to the validity of your announcements. However, we realized in 0.8.1 that users should be the once to decide whether they want to have extra ROAs or not. Therefore this clean-up should have been removed then.

This release removes this clean-up and introduces no other changes. We recommend that users who did not upgrade already upgrade to this release. However, if you already successfully upgraded to 0.8.1, then upgrading to this release is not needed.

0.8.1 'The Gentle Art'

The ROA guidance introduced in release 0.8.0 was more strict than it should be. This release allows users to create redundant ROAs once again, while providing guidance in the form of warnings and suggestions only. Full documentation on the Krill suggestions have been added to the online documentation.

In addition to this we have included some small improvements for the Krill Publication Server.

0.8.0 'The Art of ROA Maintenance'

This release includes all changes from the -rc1 and -rc2 release candidates. In addition to this the main UI (everything except the testbed pages) now includes Spanish translations.

In summary this upgrade is recommended to all Krill users. It offers improved ROA guidance support in the UI, better status reporting on interactions between your CA and its parent and repository, and a number of improvements aimed at improving resiliency.

Furthermore, we would like to draw attention once more to our testbed which allows new users to get familiar with Krill, and existing users to try out new functionality before upgrading.

0.8.0-rc2 'The Small Print'

Because of some changes we decided to have another RC release before 0.8.0, which is now planned for Monday 26 October 2020.

The reason for these changes is that while documenting the 0.8.0 release we decided that we would like to include two more small features:

• Detect and remove surplus events at start-up #332
• Add option to force recover on Krill startup #333

Issue #332 will allow Krill to recover smoothly in case the Krill process stopped in the middle of writing changes to disk. This also allows for backup strategies where the data directory is saved in its entirety from cron while Krill is running - half completed transactions will be discarded on restore.

Issue #333 should not be needed, but is added as an option in case of severe data corruption. It will force Krill to try to go back to its last 'recoverable' state. Of course this state may be too far in the past to be useful. So, please make sure you do your backups.

We have now added the updated translations for: German, Dutch, Portuguese and French. That only leaves Spanish out of the currently supported languages. Since these changes do not change the logic we feel safe to include Spanish when we release 0.8.0 without the need for an additional release candidate.

0.8.0-rc1 'Festina Lente'

As of now we will use release candidates as part of the Krill release process. If no major issues are found we plan to do the real 0.8.0 release on Monday 19 October 2020.

This new release brings a number of internal improvements, as well as new features.

New or updated features:

• added detailed ROA suggestions
• warn about ROAs which are too permissive
• support AS0 ROAs (see below!)
• allow aggregation of ROAs to lower the number of objects
• allow archiving old data in order to save space
• added a best effort recovery in case data on disk is incomplete (e.g. resulting from a full disk)
• better reporting on communication with parents and repository
• re-sync with parents and repository on start-up
• crash in case data cannot be written to disk (prevent inconsistent states)

We want to invite users to test this new version and give us feedback, in particular with regards to ROA suggestions, and so-called AS0 ROAs:

ROAs that use AS0 can be used in the RPKI to indicate that the holder of a prefix does NOT want the prefix to be routed on the global internet. In our understanding this precludes that ROAs for a real ASN for those resources should be made. Krill will therefore refuse to make AS0 ROAs for prefixes already covered by a real ASN ROA, and vice versa. Furthermore the presence of an AS0 ROA implies that announcements for covered prefixes are intentionally RPKI invalid. Therefore Krill will not suggest to authorize such announcements.

Open issues before 0.8.0:

The UI still needs translations for the updated pages. We will reach out to our translators and include these in the release. Since text changes will not affect the inner workings of Krill we believe we can do these changes without the need for an additional release candidate cycle. If you want to contribute to the translations please contact us!

0.7.4 'Multipass!'

There is no need to upgrade to this version. It was created only so that you can continue to compile Krill locally using the latest Rust compiler.

As it turns out the use of many asynchronous calls, the cool stuff which make Krill thread safe, cause the compiler to do quite a bit of work in a process called 'Monomorphization'. The latest compiler version will go on strike as a result, unless we instruct it beforehand that more work is coming its way.

0.7.3 'Slow Food'

This release fixes an issue where the BGP Ris Dump files were reloaded and checked too frequently causing high CPU and bandwidth usage.

0.7.2 'Small Bites'

This release fixes an issue where BGP RIS Dump files that were not properly retrieved would cause a thread to choke. As this can lead to lock poisoning this type of event could cause other Krill processes to stop functioning properly. All users of Krill 0.7.0 and 0.7.1 are advised to upgrade.

In addition to this German translations have been added to the UI.

0.7.1 'Sobremesa'

This release fixes the ROA migration introduced in 0.7.0. We identified an issue where the clean up of ROAs would fail because Krill tried adding explicit forms of ROAs - with max length set - before removing the implicit definitions.

0.7.0 'Escondidinho de Lagosta'

This release brings significant improvements aimed at maintaining your ROAs. For now, Krill will download aggregated BGP dumps from the RIPE NCC Routing Information Service (*) and analyse how your ROAs affect announcements seen for your resources. In future we will extend this system, so that it can use near-real-time data, or even a local feed with your own BGP information instead.

For these changes to work well we needed to do some work on cleaning up existing ROAs. Until now Krill has allowed the creation of essentially duplicate, or nonsensical ROAs, such as:

• ROAs for an ASN and prefix with and without an explicit max length matching the prefix
• ROAs for a prefix and ASN which were already permitted by another ROA.

On upgrade Krill will clean up such redundant authorizations for ROAs. For example if the following authorizations would exist:

192.168.0.0/16 => 64496 192.168.0.0/24 => 64496 192.168.0.0/16-24 => 64496

Then only this last authorization needs to be kept, the first two are also covered by it.

Before this release it was also possible to have the same authorization with, and without, using an explicit max length. For example:

192.168.0.0/16 => 64496 192.168.0.0/16-16 => 64496

Now Krill will always use an explicit max length in the definitions. Note however, that it is still best practice to use the same max length as the announced prefix length, so Krill will just set this by default if it is not specified.

*: https://www.ripe.net/analyse/internet-measurements/routing-information-service-ris

0.6.3 'Play it again, Sam'

This release addresses an issue where users with a CA that has delegated children, which in turn had performed a key roll over in the past, could not upgrade to Release 0.6.2.

Users who already successfully upgraded to Release 0.6.2 do not need to upgrade urgently. This release includes a number of fixes for minor issues, which will also be included in the 0.7.0 Release which is due in 2-4 weeks:

• krillc issues fails with Error: Unknown API method (#248)
• krillc parents help text refers incorrectly to publisher request instead of child request (#251)
• Normalize request/response krillc help texts (#252)
• krillc incorrectly reports XML as a supported output format (#253)
• Inconsistent use of "cas" in krillc bulk subcommand summary text (#254)
• Be consistent when referring to ending with a / (#255)

0.6.2 'That was even faster!'

So, as it turns out.. the code used to determine the age of snapshot files used in the previous release was not safe on all platforms. This release fixes this!

Users who upgraded to 0.6.1 and see messages like: "Creation time is not available on this platform currently" in their logs, please upgrade!

0.6.1 'That was fast!'

This release fixes an issue where the Krill Repository Server deleted RRDP snapshot files as soon as a new notification file was published. This leads to issues in case a cached notification file is served to validators.

Users who use Krill as their own Repository Server are advised to upgrade.

Users who publish at a repository provided to them by a third party (e.g. nic.br) can safely skip this release.

0.6.0 'Go with the Flow'

The most visible change in this release is that the embedded Lagosta UI now includes French, Greek and Spanish translations. But, the vast majority of the work went into making Krill use asynchronous code.

We migrated from actix-web to Hyper. Hyper is a fast, safe and fully asynchronous web framework which has a lot of momentum behind it. This change also meant that we needed to ensure that Krill itself uses safe asynchronous code whenever it connects to a remote system, like a parent or repository, or in case of the CLI the Krill API itself.

In addition to this we improved the history API to ensure that Krill will no longer use an excessive amount of history in cases where a CA has a long history. The API is still subject to change, and therefore we will only document this in future. In the meantime however, the CLI may be used to show the history of your CA.

Lagosta:

• Now includes French, Greek and Spanish translations
• Minor improvements in functionality

Krill back-end:

• Migrated from actix-web to hyper.
• Krill now uses asynchronous code where applicable.
• Krill CA history improved. (prevent server crash due to excessive memory usage)

Breaking changes:

• The API end-points for bulk operations changed to /api/v1/bulk/*
• The API end-point for CA issues moved to /api/v1/cas/{handle}/issues
• The history API changed, this is not yet stable and therefore undocumented

0.5.0 'Serve no Turf'

The most striking change in this release is the inclusion of a new front-end: Lagosta.

Lagosta 0.1 'Fritto Misto' supports the following features:

• Set up your Krill CA under an RIR/NIR parent
• Configure your CA to publish at a remote repository
• Maintain ROAs
• Internationalization (English and Portuguese)

Please talk to us if you want to contribute other languages! Many advanced features are currently available in the CLI only, but we will continue to extend the front-end functionality.

On a technical note: the front-end is based on static HTML/CSS and JS (Vue) which is served as static files to your browser by Krill. This front-end application then uses the same API back-end as the CLI.

The following features and improvements were introduced to the core Krill and CLI:

• Added option to CLI to generate a Krill config file.
• Added check for reporting status between CAs and their parents and repository
• Added simple Prometheus endpoint (/metrics)
• Disable the embedded repository by default (see docs for info)
• Added guards against using 'localhost' in non-test environments

Breaking changes:

• The error responses have been overhauled.
• Some CLI options have been changed to make naming and behavior more consistent.

For more information please have a look at Read the Docs.

We wish to thank Cynthia Revström for the great help she provided in ironing out some issues we found when setting up Krill under ARIN.

0.4.2 'Finer Things'

This release fixes a bug, and introduces minor usability improvements:

• Certain adjacent resources were encoded incorrectly (#161)
• Let users explicitly specify a repository before adding a parent (#160)
• Allow timezone to be set on the Docker container (#156)
• Improve error messaging when failing to start Krill (#155)
• Improve readability for CLI error responses (#162)
• Introduce configurable size limits for data submitted to Krill (#158)

Note that contrary to previous versions a new CA is set up without a default repository. For most users we recommend that a remote (RFC 8181) repository is used, e.g. provided by their RIR or NIR. A repository MUST be configured before a parent can be added to a CA.

0.4.1 'Fogo de Krill'

This release fixes two issues:

• Certain resource sets were handled incorrectly (#152)
• Krill should not allow impossible max length values for ROAs (#153)

We recommend that all users upgrade to this release. There were no configuration or data model changes introduced, so the binary can just be used to replace any installed 0.4.0 release.

0.4.0 'The Krill Factor'

This release focuses on stabilizing the API and internal data format, which allows upgrades to future versions of Krill without the need for complicated data migrations. We do not expect to introduce breaking changes to the API from this point forward. Please note however, that in some cases the JSON structure in API responses might be extended with additional information in new JSON members.

Overview of changes:

• Document the Krill server API using OpenAPI 3 (#148)
• Stabilize JSON API (#141)
• Better API response when a method does not exist (#146)
• Support upgrading, preserving data (#53)
• Set up automated end-to-end testing (TA-CA-ROAs-validation) (#66)
• Add config option allowing to serve RRDP from a different host (#147)
• Let Krill log to syslog (#121)
• Audit commands and errors (#142)
• Log all RFC 8181 and 6492 protocol messages (#143)

0.3.0 'The Krilling is on the wall'

This release focused on remote publication.

You can now use Krill as an RFC8181 compliant Repository Server. If you want to use it as a dedicated repository server only, you can simply do this by not configuring any CAs in that particular instance.

You can now also update your CAs to use a remote RFC8181 Repository Server. This is particularly useful if you want to outsource the responsibility of 24/7 availability of your RPKI repository to a third party.

We have also made some breaking changes to the API. So, you may have to look again at any automation you may have set previously up for release 0.2.0.

Updated documentation can be found on Read the Docs.

Two of the known issues listed under release 0.2.0 have been solved:

• CAs now do full re-syncs when publishing (solves #116)
• RIPE NCC RPKI Validator 3.1 now validates our objects (solves #115)

The next release of Krill is aimed for early December and will focus fully on stability, and the other known issues listed under release 0.2.0. But, note that there may still be small API changes in the coming release, as we are still optimizing things.

0.2.0 'Interkrillactic, Planetary'

This release focused on testing, and fixing, any issues found when running Krill under various parent CAs (Apnic, Lacnic and RIPE NCC). ROAs were tested using routinator, OctoRPKI, FORT, RIPE NCC RPKI Validator 2.x and 3.x.

Furthermore, the CLI got a big overhaul aimed at making it easier to use, especially for users who manage one CA only. You can now use ENV variables to set defaults for the Krill instance to connect to, the token, and which CA you want to operate on.

We also added the '--api' argument which will simply print out the API call that the CLI would have made, without executing it. We plan to add proper (OpenAPI) documentation for the API, but for the moment this can help to explore it.

Updated documentation can be found on Read the Docs.

Known issues:

• Despite our best efforts RIPE NCC RPKI Validator 3.1 is the only remaining RP we tested, which does not seem to like our manifests. We will look into this again later. (#115)
• There appears to be a race condition that can cause commands to be processed twice. (#64)
• Showing the full history, or logging it in case of the above condition, uses too much memory. (#112)
• The CA and publication server can be out of sync after a re-start. (#116)

Work for the next release has already started. Release 0.3 will focus on (remote) publication, and will also solve the out-of-sync issue.

0.1.0 'A View to a Krill'

This is the first version of Krill that we are testing in the real world. Please note that the API and data structures have not yet stabilized.

Features:

• Run an embedded Trust Anchor for testing purposes
• Run a CA under an embedded Trust Anchor
• Run a CA under APNIC (Lacnic, RIPE NCC and other remote parents coming soon)
• Have multiple parent CAs for one logical CA
• Have multiple child CAs, embedded or remote
• Create ROAs based on intent
• Publish locally
• API and CLI

Known issues:

• Krill does not handle concurrent updates well. See this issue.
• The UI is very basic and behind the CLI.