| Age | Commit message (Collapse) | Author |
|---|
|
Replaces the sketch (bare ifconfig alias, isolated nat line) with what we
verified end-to-end on a real VM: the alias persisted via rc.conf, a full
pf.conf modeled on a plain host's firewall, and the pfctl/kldload steps
needed to load and check it safely.
Also documents a real bug hit along the way: `nat ... -> ($ext_if)` is
ambiguous once the interface carries more than one address (the public
IP and the private jail range both live on it) - pf can silently NAT
through the wrong one. The public IP needs to be written explicitly.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
|
|
The --ip same-subnet rule from earlier is necessary but not sufficient
once a host only has one public address: jails shouldn't get public IPs
at all (only nginx needs one), but a private range aliased onto the same
interface needs its own outbound NAT rule for jail builds to reach the
internet - nothing upstream translates it the way a NAT'd virtual network
(like the dev VM) does for free. Skipping it fails the same silent way as
the out-of-subnet case: packets leave the jail, nothing answers.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
|
|
"ready" read as a status word rather than a label for what's actually a
mutable pointer to "the current recommended base image," re-tagged as it
evolves - the exact operation we did twice this session. "default" names
that role directly: it's the value --base-snapshot falls back to, nothing
more claimed.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
|
|
FreeBSD's lang/go is a real meta-port (installs as package name "go")
that depends on whichever goNNN is current and installs the go/gofmt
symlinks itself, tracked and upgradeable - `pkg install go` replaces both
the go125 install and the manual `ln -sf` steps in one command. Verified
against a real base image and rebuild: same resulting symlinks, now
pkg-owned instead of orphaned files pkg doesn't know about.
Docs updated to check for an unversioned meta-port first, falling back to
a manual symlink only when a toolchain doesn't have one.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
|
|
Replaces the single server_name string in meta.json with a domain list
(one primary, others redirect to it) and TLSDomains tracking what the
app's certificate actually covers. Certs are issued/expanded via certbot
certonly (webroot, never the --nginx plugin, since Skiff owns and
regenerates skiff.d/*.conf itself) as one SAN cert per app, matching how
a real multi-domain vhost is normally issued.
Domain/TLS changes never touch the jail - no restart, no health check -
since they only affect nginx and the certificate. `add` is safe to retry
if DNS isn't ready yet; `remove` never calls certbot at all, so it can't
fail on an ACME/network problem.
Also updates README/USER_GUIDE for the new command surface, one-time TLS
host setup (certbot register, renewal cron, reload hook), the --ip
same-subnet requirement for jails to reach the network at build time, and
the actual base-image provisioning sequence (pkg needs an explicit
bootstrap after bsdinstall; versioned port names like go125 need a
default symlink) - all found and fixed live while getting this working
end-to-end on a real FreeBSD VM.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
|
|
Each app now gets two ZFS datasets - a disposable OS layer
(zroot/skiff/appbase/<app>) cloned from base, and a persistent data
layer (zroot/skiff/appdata/<app>) holding releases/current/env/
meta.json/data.db. `skiff rebuild --app|--all` discards and reclones
the OS layer to pick up an OS/toolchain/config-template upgrade,
guarded by the same snapshot-and-rollback safety net as deploy and
config changes; the data layer is never touched.
Also fixes stale docs (builds already run via jexec inside the jail,
not on the host; --iface already auto-aliases the jail's IP, so the
manual ifconfig step and its troubleshooting entry are gone) found
while updating the same sections for this feature.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
|
|
- Extract deploy.Run's snapshot -> apply -> restart -> health-check ->
rollback-on-failure sequence into deploy.GuardedApply, shared with
cmd/skiff's applyConfigChange, which duplicated it verbatim.
- Clean up the worktree and snapshot a deploy creates when it fails
before cutover (e.g. a failing release-phase migration), so retrying
the same sha no longer needs manual cleanup.
- Make checkout and snapshotting idempotent for a same-sha retry after
a cutover that later got rolled back on a failed health check, where
the release dir and snapshot are deliberately left behind by design.
- Document the actual cause of a jail not surviving a host reboot:
FreeBSD ships no /etc/jail.conf, so boot-time `service jail start`
finds nothing to start regardless of jail.conf.d - only name-targeted
starts read jail.conf.d directly, which is why manual starts worked.
|
|
|
