* hydra/modules/sysadmin/services.scm (<goggles-bot-configuration>): New
record type.
(goggles-bot-program): New variable.
(goggles-bot-shepherd-services, goggles-bot-activation): New procedures.
(%goggles-bot-accounts, %goggles-bot-log-rotations)
(goggles-bot-service-type): New variables.
* hydra/bayfront.scm (services): Use it.
* hydra/goggles-bot.scm (%options): Append channel.
(make-filename): Accept "channel" argument.
(log-to-file): Maintain a mapping from channels to ports.
(%channels): New procedure.
(main): Join all channels.
This machine will show up as guix-ppc64le in /workers until the local
hostname is updated to match.
* doc/cuirass.org (External machines): Add sjd-p9.
* hydra/berlin.scm (services): Add it to wireguard-service-type.
* hydra/machines-for-berlin.scm (powerpc64le): Use it \o/.
This reverts commit d641115e20. The goal of this
commit was to make sure that the "guix" specification doesn't fail when there
is no powerpc64le-linux worker around.
The "guix" specification is now only supporting x86_64-linux and i686-linux
while the "guix-other-archs" specification supports powerpc64le-linux and
aarch64-linux.
This will restore powerpc64le-linux system for the "master" specification.
This option is mostly useful when using Btrfs RAID array that should
remain mountable even when one of device has a problem. Since we are
no longer using a Btrfs RAID array, remove the option.
* hydra/berlin.scm (%common-btrfs-options): Remove "degraded"
mount option.
compress-force writes fixed 512 KiB extents for any files, while
compress can write 128 MiB ones for non-compressible files, which
should greatly reduce the number of extents needed to store the NARs,
for example.
A very large number of extents can cause 'mount' to take a long time,
as is currently the case.
* hydra/berlin.scm (%common-btrfs-options): Replace "compress-force"
with "compress".
* hydra/modules/sysadmin/services.scm (cuirass-specs): Restrict guix
specification to x86_64-linux and i686-linux. Add a new guix-other-archs for
aarch64-linux and powerpc64le-linux.
Berlin has no powerpc64le-linux offload machine available which causes the
Cuirass guix evaluation to fail. Disable it for now.
* hydra/berlin.scm: Remove it.
This is no longer necessary with
45eac6cdf5c8d9d7b0c564b105c790d2d2007799 (services: Add file system
utilities to profile), which automatically takes care of that.
* hydra/berlin.scm (packages): Remove btrfs-progs and dosfstools.
* hydra/berlin.scm (%btrfs-boot-partition): New variable.
(%btrfs-ssd-uuid, %btrfs-pool-ssd): Delete variables.
(%common-btrfs-options): Delete comment.
(bootloader): Register a second efi2 target.
[file-systems]: Register %btrfs-boot-partition. Adjust device of
the /boot/efi mount point. Add a secnod /boot/efi2 mount point.
Remove anything that has to do with btrfs-ssd-ssd, which no longer
exists. Relocate /home on the SAN.
The store being on a GRUB-invisible SAN device, the kernel and initrd
should now be copied to /boot, which is hosted on two local
SSDs (Btrfs RAID1).
* hydra/berlin.scm (%copy-kernel-and-initrd): Adjust to /boot instead
of to /store. Copy the whole file name, not just its base name.
Let warnings go through.
[services]: Rename service from copy-kernel+initrd-to-/store to
copy-kernel+initrd-to-/boot.
Even if Goog^Wthe W3C decides to add ‘httpx:’ tomorrow, it's just if not
more likely to be called XTTP.
* hydra/goggles.scm (make-line-renderer): Match https? explicitly.
This enables reconfiguring the overdrive machines via 'guix deploy'.
* hydra/modules/sysadmin/overdrive.scm (overdrive-system)
[sudoers-file]: New field.
Balancing metadata is not recommended, so do not do it. Reduce
balance usage filter threshold to 5%, which is more reasonable for a
balance running often and require less free blocks.
* hydra/modules/sysadmin/overdrive.scm (btrfs-job): Rename to:
(btrfs-balance-job). Use the same snippet as used in
'deploy-node-129.scm'.
(overdrive-system) [services] <mcron-service-type>: Adjust for job rename.
* hydra/milano-guix-1.scm: Likewise.
* hydra/monokuma.scm: Likewise.
* hydra/berlin.scm (%anonip-log-files): New variable.
(log-file->anonip-service-name): New procedure.
<services>: Add 'shepherd-requirement' field to 'nginx-configuration'.
Refer to %ANONIP-LOG-FILES instead of the inlined list.
See <http://issues.guix.gnu.org/37207> for context.
* hydra/bayfront.scm (%disable-last-modified-header): New variable.
(%ten-years-of-guix-nginx-servers): Use it for both the HTTP and the
HTTPS server.
* hydra/modules/sysadmin/overdrive.scm (overdrive-system): Use 'targets'
instead of 'target' in 'bootloader-configuration', the latter being
deprecated.
Previously guix-daemon would run with --substitute-urls="https://ci.guixhttps://bordeaux.guix"; this didn't break build farm isolation though
since the key of bordeaux.guix wasn't in /etc/guix/acl.
* hydra/modules/sysadmin/overdrive.scm (overdrive-system): Add
'substitute-urls' to <guix-configuration> record.
Add a service registering the /etc/guix/machines.scm file to enable the
offload mechanism.
* hydra/modules/sysadmin/services.scm (%offload-service): New variable.
(frontend-services): Use it.
