1. Overview
  2. Source code
  3. SSH access
  4. Workflow
    1. Freeze snapshots
    2. Thaw snapshots
    3. Bump expiration date
    4. Stop tracking a distribution
    5. Freeze exception
  5. Design notes
    1. gensnapshot
    2. Garbage collection and Valid-Until
    3. APT vs. reprepro: dist names
    4. Freeze exceptions

Overview

Our time-based snapshots of upstream APT repositories are published on http://time-based.snapshots.deb.tails.boum.org/.

These are full snapshots of the upstream APT repositories we use for building Tails ISO images. They contain exactly the same set of packages as the mirrored repository. This has the advantage that some workflows are trivially handled, e.g. working on a topic branch that installs additional Debian packages; if such snapshots were not full ones, then to work on one such branch, one would need either that to have the credentials to import new packages from Debian into our own mirror or repositories (which raises the barrier for contributing), or that during some phases of Tails development the regular Debian archive is used instead of our own mirror, which feels prone to "time to QA vs. time to release" issues.

We snapshot each upstream APT repository N times a day, and without further action, each snapshot is kept for D days.

The main goal here is to be able to freeze the APT repositories used by a branch, whenever we freeze it.

A time-based snapshot's name contains:

  • an identifier of the APT repository this snapshot is about, e.g. debian, debian-security, torproject;

  • a YYYYMMDD$ID serial, $ID being an incremental decimal number formatted on two digits (01, 02, etc.).

The APT repository mirroring infrastructure publishes the name of the latest snapshot for each mirrored repository over HTTP, in the project/trace/$archive file (example). Similarly, every ISO build exports the names of the APT repository snapshots it uses (example).

The corresponding data is not critical: we can restart the whole thing from scratch if needed, without too much pain ⇒ no need to synchronize this content to the failover server; no need to back it up.

We don't bother merging mirrored APT repositories / suites into aggregated ones. It loses information, gives us more work, and brings little value.

Source code

  • tails::reprepro::snapshots::time_based class in

  • bits scattered in the main Tails Git repository (details below)

SSH access

One must configure their SSH client to connect to the APT server:

Host incoming.deb.tails.boum.org
    Port 3003

Workflow

Freeze snapshots

For example, to encode in the $RELEASE_BRANCH branch the set of time-based APT repository snapshots that shall be used during the freeze:

    ./bin/freeze-all-APT-snapshots

Thaw snapshots

For example, to encode in the devel Git branch the fact that it is not frozen anymore, that is remove the indication that a specific set of APT repository snapshots must be used:

    git checkout devel && \
    ./auto/scripts/apt-snapshots-serials thaw && \
    git commit \
        -m "Thaw APT snapshots." \
        config/APT_snapshots.d/*/serial

Bump expiration date

We set Valid-Until of time-based snapshots 10 days after they are generated. In some cases, this can be too short, and we need to manually bump Valid-Until for a given time-based snapshot.

Only release managers and sysadmins can do such operations.

Bump one specific snapshot's expiration date

To bump Valid-Until, for a given snapshot ($SERIAL) of a given archive ($ARCHIVE), so that they are valid for $DAYS_FROM_NOW days from now:

ssh reprepro-time-based-snapshots@incoming.deb.tails.boum.org \
   tails-bump-apt-snapshot-valid-until \
       "${ARCHIVE:?}" "${SERIAL:?}" "${DAYS_FROM_NOW:?}"

Bump all snapshots' expiration date

To bump Valid-Until for every snapshot used by the currently frozen branch and the corresponding builder VM, so that they are valid for $DAYS_FROM_NOW days from now:

./bin/bump-APT-snapshots-expiration-date "${DAYS_FROM_NOW:?}"

Stop tracking a distribution

After we don't need snapshots for a given Debian distribution anymore, e.g. after we release Tails based on a new Debian, we need to stop including it in new snapshots, and to remove all corresponding time-based snapshots and the packages that are not referenced anymore.

Make sure the distribution you want to stop tracking is not used for building the last published Tails release (taking into account building the corresponding Vagrant box from scratch): if it is, then cleaning up that distribution would prevent independent build reproducibility verification.

  1. In puppet-tails, remove the bits about the distribution you want to stop tracking from the reprepro configuration templates: templates/reprepro/snapshots/time_based/*/*.

  2. In puppet-tails, remove references to the distribution you want to stop tracking from the tails::reprepro::params class.

  3. Commit, push and deploy on the system that hosts our time-based APT snapshots.

  4. In the directory for each repository (e.g. debian, torproject), replace jessie to match the codename of the distribution you want to remove in the following command and run it:

     reprepro --delete clearvanished && \
 reprepro export && \
 reprepro dumpreferences \
   | grep -E '^s=jessie' \
   | awk '{print $1}' \
   | sort -u \
   | xargs -n 1 reprepro _removereferences && \
 reprepro deleteunreferenced

  5. Delete dists/$codename/ in the directory of each repository.

  6. In the directory for each repository, delete files in the logs sub-directory that were about the distribution we just stopped tracking.

Freeze exception

Grant a freeze exception

  1. Import the package you want to upgrade into our own custom APT repository, in the suite corresponding to the branch that we want to see this package in.

    For example:

        PBUILDER_OPTIONS='--basetgz /var/cache/pbuilder/base-buster-amd64.tgz' \
    TARGET_DIST='testing' \
    ./bin/import-package libgsecuredelete

    See bin/import-package for more detailed usage information.

  2. If the imported package comes from a Debian distribution whose pinning value is at least 990 in config/chroot_apt/preferences: you can stop right here. Otherwise, read on.

  3. Add a pinning entry in config/chroot_apt/preferences for the package you imported:

     Explanation: freeze exception
 Package: XYZ
 Pin: origin deb.tails.boum.org
 Pin-Priority: 999

  4. Commit:

     git commit config/chroot_apt/preferences \
     -m "Add freeze exceptions for $(dpkg-parsechangelog -SVersion)"

  5. Push to Git.

Post-release

Thaw the packages that were granted freeze exceptions on the branches where they can be fetched from a newer time-based snapshot of the repository we've initially pulled it from. Usually that's only the case on the devel branch since the stable branch still uses older time-based snapshots.

  1. For each entry in config/chroot_apt/preferences that has Explanation: freeze exception: set Pin-Priority to -1.

  2. Commit:

     git commit config/chroot_apt/preferences \
     -m "Remove freeze exceptions added for $(dpkg-parsechangelog -SVersion)"

  3. Push to Git.

Design notes

gensnapshot

We use reprepro's gensnapshot command, that basically copies a distribution, keeping references to the packages it contains.

Compared to the "snapshots as full-blown distributions + reprepro pull" option we used in our initial experiments, we are saving a lot on database size, and thus in performance, because reprepro does less tracking on snapshots, than what it does for real distributions.

The counterpart of using snapshots created with gensnapshot is that:

  • garbage collecting expired snapshots is a bit more involved, i.e. we have to do it ourselves;

  • bumping Valid-Until for a given time-based snapshot has to be done directly in dist, without any help from reprepro; so here again, we do it ourselves

None of these problems warrant going back to the other option... and having to deal with 80GB+ Berkeley DB databases.

Garbage collection and Valid-Until

We expire snapshots older than 10 days in order to save disk space, and to prevent the reprepro database from growing too much.

To ensure that garbage collection doesn't delete a snapshot we still need, e.g. the one currently referenced in the frozen testing branch, we rely on the Valid-Until field found in Release files: the way to express "I want to keep this specific snapshot around" is to postpone its expiration date; i.e. we don't differentiate "keep a given snapshot around" from "keep a given snapshot usable".

See above for how we manage Valid-Until manually.

One advantage of this design is that we don't have to regularly update Valid-Until fields, and the corresponding signatures: we only do that on a case-by-case basis, when needed. And thus, we can actually benefit from the protections offered by APT when Valid-Until fields are present, as any snapshot will expire unless we do something about it.

In practice, the main use case for keeping a given time-based APT repository snapshot around and valid is when it's being used by a release branch:

  • testing: while it's frozen, that is for 5-10 days generally;

  • stable: that's a corner case, since stable generally uses the set of snapshots frozen during the last Tails freeze; if and when we decide to manually point stable to a different set of snapshots, then we bump Valid-Until manually.

In passing, note that we ship an empty /var/cache/apt/lists/ in the ISO ⇒ modifying Release and Release.gpg files on our APT repository won't prevent the ISO build from being deterministic.

APT vs. reprepro: dist names

We need to encode in the APT sources' base URL the exact snapshot we want to use, in order to be able to pass it to lb config --mirror-*. But this doesn't match reprepro's directory structure as-is.

Thankfully this problem can be workaround'ed with some symlinks or HTTP rewrite rules. Here's how.

Let's assume:

lb config --distribution jessie
lb config --mirror-chroot \
   http://time-based.snapshots.deb.tails.boum.org/debian/2016031101/
lb config --mirror-chroot-security \
   http://time-based.snapshots.deb.tails.boum.org/debian-security/2016031102/
[...]

Which generates this APT sources.list:

deb http://time-based.snapshots.deb.tails.boum.org/debian/2016031101/ jessie main
deb http://time-based.snapshots.deb.tails.boum.org/debian-security/2016031102/ jessie/updates main
[...]

As a result APT sends HTTP requests with URLs such as:

The corresponding files in reprepro's filesystem (given that we have one reprepro instance per mirrored archive) are:

  • in Debian archive's reprepro:

    • /srv/apt-snapshots/time-based/repositories/debian/dists/jessie/snapshots/2016032401/Release, that contains Suite: jessie/snapshots/2016032401 and Codename: jessie
    • /srv/apt-snapshots/time-based/repositories/debian/pool/XYZ

  • in Debian security archive's reprepro:

    • /srv/apt-snapshots/time-based/repositories/debian-security/dists/jessie/updates/snapshots/2016031102/Release, that contains Suite: jessie/updates/snapshots/2016031102 and Codename: jessie/updates
    • /srv/apt-snapshots/time-based/repositories/debian-security/pool/XYZ

To have the above HTTP requests translate to access to these files, we use a set of HTTP rewrite rules.

Note: this works because APT only warns when the codename in the Release file doesn't match the one requested in sources.list. There's a code comment around this check, dating back from 2004, that says something like "This might become fatal in the future". We bet that if it becomes fatal some day, it will be possible to turn it back into a warning via configuration. This affects only development builds since we're not going to configure APT in the Tails ISO to point to our own snapshots of the Debian archive anyway.

Freeze exceptions

This is a new problem brought by using "frozen" snapshot of APT repositories during a Tails code freeze: some bug, that we want to see fixed in the release we are preparing, would be resolved if we pulled an upgraded package as-is from a freshest Debian APT repository. Before we could freeze APT repositories, we would have got this bugfix for free. Now we need to grant freeze exceptions.

This is similar to "Upgrading to a new snapshot", except that we want to upgrade one package only. By definition, this only affects frozen release branches (stable, testing), and topic branches based on them: all other branches use the freshest set of APT repository snapshots available.

Most of the time, a bugfix branch we want to merge into a frozen release branch doesn't need to upgrade packages from Debian, so this is a corner case for the time being. Moreover, so far we have always dealt with this problem entirely by hand, so it's not critical to provide much improved tools. What makes it tempting to improve the situation here is mostly:

  • even though freeze exceptions will remain exceptions, frozen will add one use case:

  • this will become a relatively common operation if we are based on Debian testing some day, so let's check that it's not only possible, but also reasonably easy to handle with this design (otherwise we may have to switch to more powerful tools, such as dak + britney).

To grant a freeze exception to a given package, we simply import it into our own custom APT repository, in the suite corresponding to the branch that we want to see this package in ⇒ in the general case, the upgraded package will be installed in the next Tails release.

This works because our APT pinning ranks Tails custom APT suites at the same level as the other APT sources corresponding to the current version of Debian Tails is based on, and higher than other Debian distribution (which, in passing, implies that we have to manually pin, in Git, the packages from our custom APT suites, that we want to override the ones found in other repositories regardless of version numbers):

  • if the imported package comes from Debian stable: it will be installed simply because its version is greater than the version of the same package from Debian stable; and once we have thawed the corresponding snapshot, the package can be pulled equally from any of these two sources (Debian, and our custom APT repository), until a newer version of this package is uploaded to Debian, and then the newer one will supersede the package we have in our custom APT repository;

  • if the imported package comes from another Debian distribution, that has a pinning value strictly lower than 990, such as Debian unstable: if we did nothing more, the package would be installed because its pinning (origin deb.tails.boum.org) is higher than the one from the Debian distribution we're importing it from; however, in this case we need to track this package, and to remove it from our custom repository after we have thawed the corresponding snapshot — otherwise, due to this pinning configuration, we would stick to the version of the package we have one day imported, while in most cases we want to resume tracking the version from Debian; so, we do this that way:

    1. Import the package we want to upgrade into our own custom APT repository, in the suite corresponding to the branch that we want to see this package in.

    2. Explicitly pin, in config/chroot_apt/preferences, the upgraded package we have just imported to a value higher than 990, with a proper Explanation: field; this pinning is not required at this stage, but it is one way to encode in Git the packages we have imported, which simplifies the following (clean up) step.

    3. Once the corresponding APT snapshot has been thawed, that is once the upgraded package can be fetched from a newer time-based snapshot of the repository we've initially pulled it from: make it so branches stop using the upgraded package, and resume tracking the one available in Debian. To do that, we modify the pinning entry added at the previous step, and give it a value of -1. This should be done by the release manager, immediately after a release, when they thaw the APT snapshots used for the release, and merge it into other release branches.

      Ideas for future improvements:

      • At some point a helper tool can do this automatically, assuming we always use the same Explanation: field to mark these pinning entries. (Ideally we would simply use a dedicated file under apt/preferences.d/ for freeze exceptions, but live-build 2.x doesn't support that.)

      • Ideally we would remove these imported packages from our custom APT repository at post-release time as well, so we can get rid of the -1 pinning entries, but it really needs to be done in a 100% correct order, to ensure that after all the merges we do post-release (and sometimes at other times) between release branches, the imported packages are not present anymore in any of the corresponding APT suites.