.gitbay/wiki/Admin.org
1038 lines · 54015 bytes
1#+title: gitbay admin guide
2
3One static binary (=gitbayd=), one SQLite file, bare repositories on
4disk, and the system =git=. Schema migrations run automatically on
5startup and on every admin command.
6
7* Install
8
9Build from source (=go build ./cmd/gitbayd=), install via the vanity
10module path (=go install gitbay.org/gitbay/cmd/gitbayd@latest=), or use
11a release build: =deploy/release.sh <tag>= cross-compiles reproducible
12linux/amd64, linux/arm64, and darwin/arm64 binaries with a SHA256SUMS
13manifest (CGO off, trimpath, stripped — byte-identical per commit and
14toolchain).
15
16#+begin_src sh
17install -m 755 gitbayd /usr/local/bin/
18adduser --system --group --home /var/lib/gitbay --shell /usr/sbin/nologin gitbay
19install -d -o gitbay -g gitbay -m 750 /var/lib/gitbay
20gitbayd --config /etc/gitbay/config.toml check-config
21gitbayd --config /etc/gitbay/config.toml admin secrets init
22chown gitbay:gitbay /etc/gitbay/secret.key
23#+end_src
24
25=admin secrets init= refuses when the key file already exists, so a
26reinstall on the same host should skip it — =deploy/install.sh= does
27this with a file check before running it.
28
29=deploy/= in the source tree has a cloud-init file, a hardened systemd
30unit, and a nightly backup timer. Run as the unprivileged =gitbay= user;
31the unit's =AmbientCapabilities=CAP_NET_BIND_SERVICE= covers ports
3222/80/443 without root.
33
34** The SSH port decision
35
36- =ssh.mode = "embedded"= (default): gitbayd itself listens, normally on
37 22 — move the host's admin sshd to another port. Remotes read
38 =git@host:owner/repo= with no port gymnastics.
39- =ssh.mode = "system"=: the host sshd owns 22 and invokes gitbayd via
40 =AuthorizedKeysCommand=:
41 #+begin_example
42 AuthorizedKeysCommand /usr/local/bin/gitbayd --config /etc/gitbay/config.toml authorized-keys %t %k
43 AuthorizedKeysCommandUser gitbay
44 #+end_example
45 sshd requires that binary to be root-owned and not group/world
46 writable. Unknown keys fail authentication inside sshd, so system mode
47 requires =registration.mode = "closed"= (check-config enforces this).
48 Under the forced command the CLI's leading =--term=<cols>[,color]=
49 argument works as is; a client that sets =GITBAY_TERM= instead needs
50 =AcceptEnv GITBAY_TERM= in =sshd_config=, since that env request is
51 handled by the host's sshd, not gitbayd.
52
53* Configuration reference
54
55=/etc/gitbay/config.toml=. =check-config= validates and names every
56contradiction; =--no-host-checks= skips port/path probes.
57=gitbayd admin config show= prints the configuration in effect as TOML,
58every default filled in and =smtp_pass= redacted. A file that fails
59validation still prints, followed by the contradiction.
60
61** [server]
62- =root= (default =/var/lib/gitbay=) — repositories, database, host
63 keys, ACME cache all live here.
64- =site_url= (required) — canonical =https://host=; drives ACME, clone
65 URLs, mail links.
66- =secret_key_file= (default =/etc/gitbay/secret.key=) — the keys that
67 seal CI secrets, webhook secrets, mirror tokens and push device
68 tokens in the database. Must be outside =root=, mode 0600, readable
69 by the daemon's user. See "Secret key" below.
70- =source_repo= (optional, =owner/name=) — the repository this instance
71 develops itself in. Startup warns when the running build's commit is
72 not on that repository's default branch, which is how a binary built
73 from an unmerged branch stops being invisible. Leave it unset unless
74 the instance hosts its own source.
75
76** [ssh]
77- =mode= — =embedded= | =system= (above).
78- =port= (22) — embedded listener port.
79- =host_keys= — list of private key paths; empty generates an ed25519
80 key at =<root>/ssh/host_ed25519=.
81
82** [http]
83- =addr= (=:443=), =tls= — =acme= | =files= | =off=.
84- =acme=: certificates via TLS-ALPN-01 on the HTTPS port, cached at
85 =<root>/acme=; =acme_email= for the CA account; =acme_http_addr=
86 (=:80=, ="off"= to disable) adds HTTP-01 and an https redirect —
87 failing to bind it is a warning, not fatal. Requires an =https://=
88 site_url with a public DNS name.
89- =files=: =cert_file= + =key_file=.
90- =off=: plain HTTP — development, or behind a TLS-terminating proxy.
91- The HTTPS listener accepts TLS 1.2 and 1.3 only (=serverTLS= in
92 =cmd/gitbayd/tls.go=), with the default cipher suites of the Go
93 release the binary was built with. =openssl s_client -connect
94 <host>:443 -tls1_1= fails the handshake.
95
96=trusted_proxies= lists the addresses or CIDRs of reverse proxies in
97front of the daemon. A request from one of them is attributed, for API
98rate limiting, to the last =X-Forwarded-For= hop that is not itself a
99trusted proxy; from anyone else the header is ignored. Empty, the
100default, is right when gitbayd terminates TLS itself.
101
102A reverse proxy in front of gitbay must not buffer responses, or the
103build page's live log arrives only when the build ends;
104=X-Accel-Buffering: no= covers nginx.
105
106** [web]
107- =mode= — =view_only= (default) | =accounts=. In view_only the mutating
108 web routes are never registered; in accounts, browser sessions are
109 minted over SSH (=web login=), and users with write access can create
110 repos, comment, and make simple file edits (which commit unsigned,
111 honestly). =password_auth= is reserved and currently rejected.
112- =title= — the instance's display name in the rail and page titles;
113 empty falls back to the site host. Lower case is the convention for
114 gitbay itself.
115- =privacy_notice= — operator text shown on =/privacy= under the fixed
116 statement.
117
118** [registration]
119- =mode= — =closed= (default) | =invite= | =open=. invite/open require
120 [mail]. See the user guide for the flows.
121
122- =pending_expiry= (empty, never) — a duration such as ="168h"=; a
123 self-registered account still unverified after that long is removed,
124 hourly and at start, audited as =pending.expired=.
125
126- =notify_admin= (false) — mail every instance admin when an account
127 becomes active: an invite redeemed, or an open-mode signup that
128 verified its address. The unverified row an open signup creates is
129 not reported, because anyone can post the form and mailing on that
130 would point a flood at the admins. Recipients are the verified
131 primary addresses of active admins who have activity mail on, the
132 same rule any other notice follows, so an admin with no verified
133 address hears nothing. The notice is queued, so a dead SMTP host
134 shows up in the admin page's Mail table instead of failing the
135 registration. Requires [mail].
136
137** [mail]
138- =smtp_host= (host:port; 587 assumed, 465 with =tls = "implicit"=),
139 =from=, optional =smtp_user= / =smtp_pass=. Required for invite/open
140 registration and self-service =email add=; in closed mode you may omit
141 it entirely and assert addresses by hand (below).
142- =tls= — =starttls= (default) or =implicit= (TLS from the first byte,
143 for relays on 465).
144- =require_tls= — with =starttls=, a relay that does not offer STARTTLS
145 gets no mail: delivery fails and retries, and the admin page's Mail
146 table shows the error. Unset, it is on for every relay except
147 =localhost= and loopback addresses; set =false= to allow plaintext
148 to a remote relay.
149
150** [push]
151Push notifications to Apple devices, delivered by gitbayd talking to
152APNs directly over HTTP/2, authenticated by an ES256 JWT signed with an
153operator-supplied provider key. Off unless configured.
154
155- =enabled= (false).
156- =key_file= — path to the =.p8= provider key from Apple's developer
157 portal (Certificates, Identifiers & Profiles → Keys). It belongs at
158 =/etc/gitbay/apns.p8=, mode 0600, owned by the account gitbayd runs
159 as. Read and validated at startup: it must parse as a PEM-wrapped
160 PKCS#8 EC (P-256) private key, or the daemon refuses to start rather
161 than fill a queue nobody is watching.
162- =key_id=, =team_id= — the key's id and your Apple developer team id,
163 both from the same portal page.
164- =topic= — the app's bundle identifier. *An APNs key belongs to a
165 bundle ID.* gitbay.org pushes to the App Store build under its own
166 bundle id; a self-hoster who wants push ships their own iOS build
167 under their own bundle id, with its own =.p8= key from their own
168 developer account, and points =topic= at that id. There is no way to
169 push to someone else's build, by design — this is Apple's model, not
170 gitbay's.
171- =environment= — =production= or =sandbox=, naming the APNs host
172 rather than taking a URL, so a typo cannot aim the key at a host that
173 is not Apple's.
174
175All five of =key_file=, =key_id=, =team_id=, =topic= and =environment=
176are required when =enabled= is true; validation runs at config load,
177so a misconfigured =[push]= is caught before the daemon serves
178anything. The delivery queue (a device's undelivered and attempted
179pushes) is capped the same way the mail queue is, by =[retention]
180push=.
181
182** [backup]
183- =age_recipients= (optional) — age public keys (=age1...=). When set,
184 =admin backup= encrypts every archive to them and appends =.age= to
185 its name. Generate the pair off the host with =age-keygen=; only the
186 public key goes here, so the host writes archives it cannot read.
187 The restic copy is unaffected: the offsite job stages its own
188 =VACUUM INTO= of the live database and snapshots =/var/lib/gitbay=,
189 not the archives.
190
191** [api]
192- =enabled= (false) — the JSON API surface; see [[API]]. Off
193 means no credential-bearing HTTP endpoint exists at all.
194
195** [webhooks]
196- =allow_local= (false) — permit webhook and mirror targets on
197 loopback, private, shared (100.64.0.0/10), link-local or multicast
198 addresses. Leave off unless you know why you need it (SSRF).
199
200** [limits]
201- =clone_timeout= (3600s) — cap on =repo import= fetches.
202- =max_blob_bytes= (100MB) — cap on raw file serving over the web.
203- =max_asset_bytes= (512MB) — cap per uploaded release asset.
204- =max_snippet_bytes= (1MB) — cap per snippet file.
205- =max_snippets_per_user= (0, unlimited) — snippets an account may own.
206- =max_repos_per_user= (0, unlimited) — repositories an account may own
207 directly; =repo create=, =fork= and =import= refuse past it.
208 Organizations are not capped.
209- =max_bytes_per_user= (0, unlimited) — disk the account's own
210 repositories may take; a push may be no larger than what is left.
211- =pack_concurrency= (3), =pack_per_principal= (2), =pack_queue= (32),
212 =pack_queue_wait= (="60s"=) — git pack generation (clones, fetches,
213 =git archive --remote=, web archive downloads) over SSH, smart HTTP
214 and git:// shares one
215 budget: this many at once, this many per account (per client
216 address when anonymous: an IPv4 address, or an IPv6 /64), and this
217 many waiting for at most the wait. Anonymous clients together hold
218 at most =pack_concurrency= − 1 slots when it is above 1: an account
219 (SSH key, bearer token or web session) can take the last slot when it
220 is free, and anonymous clients cannot hold it.
221 Past that an SSH client gets "the server is busy…" and exit 1, HTTP
222 gets 503 with =Retry-After: 30=, git:// an =ERR= line; the daemon
223 logs a =pack limit= warning naming the transport and whether the
224 client was signed in, at most once a minute per transport. An SSH client
225 that disconnects while queued leaves the queue; an HTTP or git:// one
226 keeps its place until the wait runs out. A running clone is killed
227 when its client disconnects, or when no write to the client completes
228 for two minutes: a client reading below about 550 B/s, or an HTTP
229 request body that takes over two minutes with nothing written back,
230 is cut. Ref listings (info/refs, protocol v2
231 =ls-refs=), pushes and =repo download= are outside the budget. For the
232 three counts 0 means the default and a negative value turns that
233 bound off. The defaults suit a four-core host; see [[Performance]].
234 With =ssh.mode = "system"= each SSH session is its own process and
235 SSH clones are not counted.
236- =max_pack_bytes= (2 GiB) — the largest pack one push may send,
237 enforced as =receive.maxInputSize= and lowered to what an owner's
238 storage quota has left.
239- =ssh_auth_rate= (10) — SSH authentication failures per client address
240 per minute on the embedded listener; see below.
241
242** [git_daemon]
243- =enabled= (false), =port= (9418) — the anonymous =git://= listener.
244 Serves only public repositories that additionally ran
245 =repo settings git-daemon <repo> on=.
246
247** [mirrors]
248- =pull_interval_minutes= (15) — how often pull mirrors fetch their
249 upstream. Push mirrors sync shortly after each local ref update.
250 Mirror URLs pass the same SSRF rules as webhook targets, when saved
251 and again before every sync; git then connects only to the addresses
252 that were checked (=http.curloptResolve=) and does not follow
253 redirects, so a mirror of a renamed repository fails until its URL
254 is updated. Needs git 2.37 or later on the server; with an older git
255 the worker logs an error at start and syncs no mirror, recording the
256 reason on each. Sync ignores the system and global gitconfig.
257
258** [go_import]
259Vanity Go module paths, one per line: ="host/module" = "owner/repo"=.
260Requests with =?go-get=1= at or under the module path answer with the
261go-import meta tag pointing at the repository's HTTPS clone URL, so
262=go install host/module/cmd/...@latest= resolves. The repository should
263be public (the module path itself confirms it exists).
264
265* Users, email, invites
266
267Every =gitbayd admin= subcommand except =backup=, =gc= and the one-shot
268backfills is a wrapper that dispatches the registry command of the same
269name as the host: an admin context with no account behind it, so its
270audit rows carry no actor and =source: host=. The same commands run in an
271instance admin's SSH session (=ssh git@<host> admin ...=) and write the
272same rows with the key fingerprint as source. One implementation, two
273credentials.
274
275#+begin_src sh
276gitbayd admin user create alice --key alice.pub --email a@example.org --verified [--admin]
277gitbayd admin email verify alice a@example.org # admin assertion, no SMTP needed
278gitbayd admin invite --email b@example.org # mails a code; prints it if no SMTP
279#+end_src
280
281"Verified" means SMTP-confirmed or host-admin-asserted; the database
282records which. Verified emails are what make commit signatures
283meaningful — an unverified address never produces a =verified= badge.
284
285* Audit and account control
286
287The audit log is the security feed (events are the product feed): every
288successful mutating command with its argv and source credential (SSH key
289fingerprint or API), every refused one (exit 3 or 4) as =refused
290<command>=, refused pushes as =refused git-receive-pack= (the access
291check) or =refused push= (a branch or tag rule, a release anchor or an
292unsigned commit, with the repository and ref names), hook socket
293requests failing the peer or push-token check as =refused hook= (never
294the token), registrations, admin actions, force-pushes, and auth
295failures/throttling. A refusal row
296keeps the flag names and the first positional, not the values.
297Refusals are recorded up to ten a minute per account and 600 a minute
298across the instance; past either, one =refused.throttled= row stands
299for the rest of that minute. The caps bound the embedded listener, the
300web and the API; under =ssh.mode = "system"= each =gitbayd shell=
301connection counts separately. Secrets never appear — they travel on
302stdin, never in argv.
303
304Each row carries the SHA-256 of the row before it. =gitbayd admin audit
305verify= opens the store as other admin commands do, applying pending
306migrations, so run it with the binary that matches the daemon. It
307recomputes the chain and exits 1 naming the first row that was
308edited or whose predecessor was removed. The chain is unkeyed: whoever
309can write the database can recompute every hash after an edit, and
310verify then finds nothing. It catches an edit only when the later
311hashes were not recomputed. Retention removing the oldest
312rows is not a break. Rows written before the chain existed are counted
313and skipped; when every row is such a row, verify warns and exits 1,
314since clearing the hash columns looks the same. After an upgrade that
315clears with the first new audit row.
316
317Removing the newest rows leaves no break, and neither do rows written
318afterwards under the freed ids. The database cannot show either. The
319daemon logs every row it writes to its journal, outside the database
320(=journalctl -u gitbayd -g 'INFO audit '=), and verify prints the last
321id and hash: comparing them with the newest journal line is the check
322for any change, recomputed hashes included. Rows written by
323host =gitbayd admin= commands, and by =gitbayd shell= when =ssh.mode =
324"system"=, are not copied to the journal.
325
326#+begin_src sh
327gitbayd admin audit [--actor u|-] [--action prefix] [--since 24h|7d|date] [--limit n] [--json]
328gitbayd admin audit verify # check the hash chain; exit 1 names the first bad row
329ssh git@<host> audit ... # the same, from an admin session
330ssh git@<host> admin user list [--state active|pending|disabled|admin]
331ssh git@<host> admin user show <name> # keys, emails, orgs, tokens, sessions
332ssh git@<host> admin user limits <name> [--repos n|default] [--bytes n|default] # per-account caps
333ssh git@<host> admin user promote <name> # grant instance admin
334ssh git@<host> admin user demote <name> # remove it; the last admin is refused
335gitbayd admin user promote <name> # host-local: recovery when no admin key is reachable
336gitbayd admin user disable <name> # suspend: SSH, web, API all refused;
337gitbayd admin user enable <name> # sessions dropped, nothing deleted
338gitbayd admin user delete <name> --yes # only for accounts anchoring nothing:
339 # refused (with each blocker named) while
340 # the account owns repos, authored
341 # issues/MRs/comments/reviews, or is an
342 # org's only admin
343#+end_src
344
345=--actor= takes a username, or =-= for rows with no actor: host commands
346and auth failures. =--action= is a prefix, so =cmd repo= catches every
347repository command and =admin= every host or admin-session action.
348=--since= is a duration back from now (=30m=, =24h=, =7d=) or a date.
349
350=admin user list= pages by username (=--limit=, =--cursor=) and carries
351each account's state and =last_seen=, the newest use of any of its SSH
352keys or API tokens. =admin user show= adds the keys with their last use,
353each address with how it was verified, PGP keys, org roles, the owned
354repository count, API token names, and live browser sessions. Both are
355Both are refused to non-admins, like =audit=, on every surface.
356
357=/admin/users= is the same list in a browser, linked from the admin
358page: the state filter the command takes, keyset paging on its cursor,
359and a row per account with promote, demote, disable and enable, each
360dispatching the command. Demote and disable ask for the username to be
361typed, since both take someone's access away. Creating and deleting an
362account, issuing an invite and asserting an address stay on the command
363line: each takes a key, mints a credential, or cannot be undone. A
364non-admin gets the 404 a missing page would, so the URL confirms
365nothing.
366
367Promotion needs an active account: a pending or disabled one is refused.
368Demotion is refused when it would leave no admin, over SSH, in the
369browser and on the host alike, so the host-local =promote= is the way
370back in when the only admin key is lost.
371
372Instance admin carries no right on anyone's repository: policy does not
373consult it, and a private repository still answers not-found to an
374admin. Moderation goes through explicit overrides that skip the access
375check and write their own audit row:
376
377#+begin_src sh
378ssh git@<host> admin repo list [--owner o] [--visibility public|private] # size, last push
379ssh git@<host> admin repo archive|unarchive <owner/name>
380ssh git@<host> admin repo visibility <owner/name> public|private
381ssh git@<host> admin repo delete <owner/name> --yes
382#+end_src
383
384Each lands in the audit log as =admin repo.<action>= naming the
385repository, on top of the =cmd= row every mutating command gets.
386
387=limits.ssh_auth_rate= (10) throttles per-IP authentication *failures*
388per minute — successful auths never count and clear the slate.
389=limits.max_pack_bytes= is enforced as =receive.maxInputSize= on every
390push.
391
392=limits.write_rate= (60) bounds *mutating commands per account per
393minute*. It is counted in the dispatcher, so SSH, the JSON API and the
394web spend one budget and a caller cannot refresh it by changing surface;
395=limits.api_rate= stays in front of it, bounding a network source rather
396than an account. A command is one token whatever it writes, so a bundle
397import costs one and only a loop of separate commands spends the budget.
398Read-only commands, the runner protocol (a build streams its log in many
399small writes) and the host CLI are exempt. Refusals exit 4 and say when
400to retry. A negative value turns the limit off; it matters most with
401=registration = "open"=, where every write also queues notification mail
402and webhook deliveries.
403
404* Queues
405
406Every background worker keeps a backlog and a failure state. An instance
407admin reads them all in one place:
408
409#+begin_src sh
410gitbay dashboard --json | jq .queues # webhooks, mail, push, mirrors, builds, deps
411#+end_src
412
413Per worker: pending, retrying (pending with a failed attempt) and
414dead-lettered counts with the oldest pending age, and the retrying or
415failed rows themselves, capped at twenty each. Builds list what is
416running and then what is pending, each since when, so a build no runner
417is scoped to claim is visible here rather than only in its repository;
418mirrors list the ones whose last sync failed;
419dependency checks list the ones whose last check errored. Non-admins get
420no =queues= key at all.
421
422Push rows name the device id, never the token. Watch this one after
423configuring =[push]=: a =key_id= or =team_id= Apple did not issue passes
424config validation, which can only check that the =.p8= parses, and then
425every send comes back =403 InvalidProviderToken= and dead-letters on its
426first attempt.
427
428In accounts mode the same read renders at =/admin=, linked from the rail
429for admins. Anyone else gets a 404 there.
430
431A dead-lettered mail is logged as =notification dead-lettered mail=<id>=,
432with the address redacted out of the relay's error. The id is the queue
433row: find it in the Mail table on =/admin=, or in =dashboard --json=,
434where the recipient and the unredacted error are. That is deliberate —
435see the Threat-Model page.
436
437* Maintenance
438
439#+begin_src sh
440gitbayd admin stats [--json] # counts, database size, per-repo disk
441ssh git@<host> admin stats [--json] # the same, from an admin session
442gitbayd admin gc [--repo owner/name] # git gc: repack and prune; per-repo sizes
443gitbayd admin gc --aggressive # thorough repack; slow, rarely needed
444gitbayd admin gc --lfs # also drop LFS objects no pointer names (older than a day)
445#+end_src
446
447A history rewrite leaves the commits it removed reachable through
448=refs/merge-requests/N/head= of the merge requests that landed them, so
449they stay fetchable by anyone who can read the repository. Nothing drops
450a head ref on its own — an open or source-gone MR is merged through it,
451and a merged or closed one keeps its diff readable through it — so the
452cleanup is a command an instance admin runs, naming the MRs:
453
454#+begin_src sh
455ssh git@<host> admin mr prune owner/name 1 2 3 --yes
456#+end_src
457
458It refuses an open or source-gone MR, deletes the named refs, runs
459=git gc --prune=now= on that one repository so the objects go at once
460rather than after git's two-week grace, leaves a system comment on each
461MR, and audits as =admin mr.prune=. The MR keeps its title, comments,
462reviews and head sha; =mr diff= and the MR page say the head is gone.
463Run it when nothing is pushing to that repository: without the grace, a
464push caught between leaving quarantine and writing its ref loses its
465objects. Objects also survive in offsite backups until those are
466rewritten; see "Removing a repository's history from every snapshot".
467
468=deploy/cloud-init.yaml= ships a =gitbay-gc.timer= that runs =admin gc=
469weekly (Sunday 07:00 UTC). Imported repositories keep whatever pack
470layout the source sent, so a first manual =admin gc= after a bulk
471import is worthwhile.
472
473* Backup and restore
474
475#+begin_src sh
476gitbayd admin backup --out /var/backups/gitbay/backup.tar.gz
477gitbayd admin backup --verify /var/backups/gitbay/backup.tar.gz # read it back
478#+end_src
479
480One archive: a consistent SQLite snapshot (taken *before* the
481repositories are read, so the database never references objects the
482archive missed), every repository, and the SSH host keys. Excluded:
483hook socket, regenerated hook scripts, WAL files. Safe to run against a
484live daemon. =--out= must be outside =server.root=, or the next full
485backup would carry the archive. Each run removes the snapshot
486directories (=.gitbay-snap-*=) and temporary archives (=.*.tmp-*=) a
487killed run left beside its archive once they are a day old, and prints
488each one it removes.
489
490=--verify= reads an archive back: the snapshot must pass SQLite's
491integrity check, every repository the snapshot names must be in the
492archive, and each must pass =git fsck --connectivity-only=. It
493extracts the repositories to a temporary directory for that, so it
494needs free space the size of the repositories. A database-only archive
495is checked for integrity and says so. Exit is non-zero on damage, a
496missing repository or a missing object.
497
498A full backup holds =<root>/backup.lock= from its database snapshot to
499its last repository. While it runs, =repo delete=, =repo rename=,
500=repo transfer=, =admin repo delete=, =org rename=, =admin mr prune=
501and =gitbayd admin gc= refuse with "a backup is running"; retry when it
502finishes. Database-only backups take no lock. A pack that git's own
503automatic gc removes during the walk is skipped; each repository's refs
504are archived before its objects, so the refs still find their objects,
505and =--verify= reports it if one does not.
506
507Verifying an archive of unknown origin: run it as an unprivileged
508user. The connectivity check runs git with =--git-dir= on each
509extracted repository, so a directory that is not a repository fails
510rather than git checking an enclosing one. =objects/info/alternates=
511and a =commondir= directly in a =*.git= directory are not extracted,
512so an archive cannot use either to have git read another repository's
513objects or refs on the host. Git still reads each archived
514repository's own =config=.
515
516Archives carry a directory entry for every directory, including an
517empty one, so a bare repository whose refs are all packed restores as
518a repository. Archives written before this release do not: extracting
519one can leave a repository's =refs/= directory missing, which stops
520git from recognizing it as a repository at all. =gitbayd admin backup
521--verify <archive>= names the repositories this affects; the fix is
522=mkdir -p <root>/repos/<owner>/<name>.git/refs= for each one, after
523which it opens normally.
524
525With =[backup] age_recipients= set the archive is =<name>.tar.gz.age=
526and =--verify= needs the private key:
527
528#+begin_src sh
529gitbayd admin backup --verify gitbay-20260927-090000.tar.gz.age --identity ~/.config/gitbay/backup-identity.txt
530age -d -i ~/.config/gitbay/backup-identity.txt gitbay-20260927-090000.tar.gz.age | tar -xz -C /new/root
531#+end_src
532
533The identity lives off the host (with the secret key file and the
534restic credentials), so verifying an encrypted archive happens there
535or on a restore host.
536
537Restore: extract into an empty directory, point =server.root= at it,
538restore =server.secret_key_file= from its own copy (mode 0600, owned
539by the account gitbayd runs as), then start gitbayd. No archive carries
540the key file, and without it gitbayd refuses to start. Host keys are
541preserved, so clients keep their known_hosts entries; hooks regenerate
542at startup.
543
544** Schedule and recovery point
545
546Two timers, because the two halves of the data have different exposure.
547
548- =gitbay-backup.timer=, nightly. The full archive above, last 7 kept.
549- =gitbay-db-backup.timer=, hourly. =admin backup --db-only=, which
550 writes the SQLite snapshot alone, last 48 kept. A few MB against the
551 full archive's hundreds, which is what makes the frequency affordable.
552
553The split follows what a loss would actually cost. Repositories are git,
554so a mirror or any clone is a second copy; the database is the only copy
555of issues, merge requests, comments and review state. So the recovery
556point is about an hour for the data that exists nowhere else, and a day
557for the data that does.
558
559Continuous replication (litestream and similar) was considered and not
560adopted. It would take the database's recovery point to seconds, but the
561repositories would still be on the nightly archive, so a restore could
562produce a database referencing commits the repository backup does not
563have. Consistency between the two halves is worth more here than latency
564on one of them. Revisit if repository replication becomes continuous
565too.
566
567** Offsite copies
568
569bay1 also takes a nightly restic snapshot of =/var/lib/gitbay= and
570=/var/lib/gitbay-stage= (a database copy and =config.toml=, staged
571there) to an S3 bucket at Scaleway, with a key that can only add
572snapshots. Nothing else under =/etc/gitbay= is in it: not the secret
573key file, not =apns.p8=. The key that can
574remove them lives on the operator's machine, in
575=~/.config/gitbay/offsite.env=, and never on bay1: a compromised host
576cannot destroy its own history. Forgetting, pruning and rewriting all
577run from there.
578
579*** Removing a repository's history from every snapshot
580
581A history rewrite plus =admin mr prune= takes commits off the server,
582but every snapshot taken before it still holds them, and the retention
583window is the only thing that ages them out. To remove them now,
584rewrite the snapshots without that repository rather than forgetting
585the snapshots: everything else in them stays restorable. The next
586nightly run adds the repository back in its current state.
587
588Repositories are stored under the name they had on disk when each
589snapshot was taken, so a renamed repository needs every name it has
590carried. Check what an older snapshot holds before choosing the paths:
591
592#+begin_src sh
593set -a; . ~/.config/gitbay/offsite.env; set +a
594restic $RESTIC_OPTS snapshots
595restic $RESTIC_OPTS ls <old-snapshot> /var/lib/gitbay/repos/<owner>
596#+end_src
597
598Then dry-run, apply, prune, and confirm nothing matches:
599
600#+begin_src sh
601EXCL="--exclude /var/lib/gitbay/repos/<owner>/<name>.git --exclude /var/lib/gitbay/repos/<owner>/<old-name>.git"
602restic $RESTIC_OPTS rewrite --dry-run $EXCL # "would modify N snapshots"
603restic $RESTIC_OPTS rewrite --forget $EXCL # new snapshots replace the originals
604restic $RESTIC_OPTS prune # drops the data nothing references
605restic $RESTIC_OPTS find <name>.git <old-name>.git # expect no output
606#+end_src
607
608=--forget= is what makes the originals go; without it the rewritten
609snapshots sit beside them and the data stays referenced. Snapshot IDs
610change; their times do not. Done for krz/keycask (formerly rust-pass)
611on 2026-09-18, across 22 snapshots.
612
613** Secret key
614
615CI secrets, webhook secrets, mirror tokens and APNs device tokens are
616stored sealed: AES-256-GCM under a key in =server.secret_key_file=,
617each value prefixed with the id of the key that sealed it
618(=gbs1:<id>:=). The key file is not in the database, not under
619=server.root=, and therefore in neither the local archives nor the
620main restic repository. It must be copied off the host separately;
621without it a restored database's secrets cannot be opened, and
622gitbayd refuses to start against them. A separate restic repository
623for it and =apns.p8= is planned (runbook D of the data-at-rest plan)
624and not yet in place, so today the only off-host copy is one the
625operator makes by hand after =init= and after every =rotate=.
626
627#+begin_src sh
628gitbayd admin secrets init # once; deploy/install.sh does it on first install
629gitbayd admin secrets check # open every value, count by key
630gitbayd admin secrets rotate # new key, reseal, retire the old one (as root)
631#+end_src
632
633- Missing file: every gitbayd process that opens the database refuses
634 to run and names the path, including =serve= and, in system mode,
635 =authorized-keys=. =migrate= does not need it.
636- Backups: =admin backup= and =admin backup --verify= do not need the
637 key file; a restore does.
638- Wrong key: =serve= stops at startup naming the first row that does
639 not open; =secrets check= does the same without starting anything.
640- Upgrade: the first start after the upgrade seals every value still
641 in clear and logs =sealed secret values=.
642- Rotation: =rotate= adds a key, reseals every value under it in one
643 transaction, then removes the old keys. Run it as root, since it
644 replaces the key file in =/etc/gitbay=; the file keeps its owner.
645 The daemon re-reads the file when it changes, so it needs no
646 restart. Copy the new file off the host afterwards. =init= and
647 =rotate= hold an flock on =<key file>.lock= while they run, so a
648 second run waits for the first.
649- Push devices are looked up by the SHA-256 of their token
650 (=push_devices.token_hash=), since two seals of one token differ.
651
652** Restore drill
653
654A restore onto a clean host, run quarterly and after any change to the
655backup code (=cmd/gitbayd/backup.go=, the offsite job), and recorded
656below. The disaster it rehearses is losing bay1, so the local archives
657are gone with it and the sources are the main offsite restic
658repository (repositories, LFS, the staged database, =config.toml=),
659the off-host copy of =secret.key= and =apns.p8= (a keys repository once
660runbook D creates it; until then the operator's hand-made copy), and
661the operator's password manager (=offsite.env=, the keys repository's
662password and token once it exists, =backup-identity.txt=). The steps are in the data-at-rest
663plan's operator runbook
664(=docs/plans/2026-09-27-data-at-rest-and-backup.md=).
665
666Time to service runs from the clean host's first root login to the
667first successful =git clone= over SSH from it. The recovery point is
668the time of the newest restic snapshot restored.
669
670No drill has been run yet; the procedure above is written but
671unexercised, and #259 stays open until the first row below is
672recorded.
673
674| Date | Host | Snapshot restored (UTC) | Time to service | DB integrity | Connectivity | LFS | Release assets | Host key | Secrets | Notes |
675|------+------+-------------------------+-----------------+--------------+--------------+-----+----------------+----------+---------+-------|
676
677* Upgrades
678
679Replace the binary, restart the unit. Migrations apply automatically and
680are transactional; hook scripts under =<root>/hooks= are rewritten at
681startup to point at the current binary path.
682
683* CI runner
684
685=gitbay-runner= executes builds queued by pushes and merge requests. It
686polls over SSH with a key of scope =runner=, which reaches only the
687runner protocol and read-only git (a runner executes arbitrary
688repository code, so the key it holds must not do more). A runner key
689claims builds only for the repositories it is attached to, by =repo
690runner add= from a repository admin or an instance admin; an admin key
691claims any. Users attach their own runners: see the Users page. For an
692instance runner, run it as a dedicated unprivileged user on a non-admin
693account. =admin user create --key= registers a full-scope key, so the
694runner key is added afterwards through a bootstrap key that is then
695removed, and attached to each repository it should build:
696
697#+begin_src sh
698useradd --system --create-home --home-dir /var/lib/gitbay-runner ci-runner
699sudo -u ci-runner ssh-keygen -t ed25519 -N "" -f /var/lib/gitbay-runner/.ssh/id_ed25519
700ssh-keygen -t ed25519 -N "" -f /tmp/ci-bootstrap
701gitbayd --config /etc/gitbay/config.toml admin user create ci --key /tmp/ci-bootstrap.pub
702ssh -i /tmp/ci-bootstrap git@127.0.0.1 keys add --scope runner < /var/lib/gitbay-runner/.ssh/id_ed25519.pub
703ssh -i /tmp/ci-bootstrap git@127.0.0.1 keys remove "$(ssh-keygen -lf /tmp/ci-bootstrap.pub | awk '{print $2}')"
704rm /tmp/ci-bootstrap /tmp/ci-bootstrap.pub
705gitbay-runner -remote git@127.0.0.1 -workdir /var/lib/gitbay-runner/work
706#+end_src
707
708#+begin_src sh
709gitbay repo runner add krz/site < /var/lib/gitbay-runner/.ssh/id_ed25519.pub
710#+end_src
711
712=-jobs N= runs N builds at once. Claiming is one transaction that
713selects and updates, and each build works in its own =build-<id>=
714directory, so workers do not collide; idle polls are staggered across
715the interval so N of them do not wake together. The drop-in's weights
716below are per service, not per build, so raising =-jobs= divides them
717rather than multiplying the host's load.
718
719=admin runners= shows which account each runner polls as, and what each
720may claim. A runner key claims builds only for the repositories it is
721attached to: with none attached it claims nothing, and =-repos= may only
722narrow within them. An admin's full-scope key claims any repository —
723that is what =-repos= was for — and still works for the protocol during
724a rotation. A merge request head from a fork is built in the target
725repository as untrusted: the claim carries no secrets, and only a runner
726started with =-untrusted= takes it. Same-repository heads were built by
727their branch push and are not built again.
728
729=make deploy-runner= also installs
730=deploy/gitbay-runner.override.conf= as a systemd drop-in: =Nice=10=,
731=CPUWeight=30=, =IOWeight=30=, so a build never starves the host's sshd,
732the daemon or the backup timers, and =NoNewPrivileges=,
733=ProtectSystem=full=, =ProtectKernelTunables=, =ProtectControlGroups=
734and =RestrictSUIDSGID=, so a step cannot reach outside its workspace
735and the runner's home. The e2e suite alone starts sixty
736daemon instances; without the drop-in a deploy's copy over the admin
737sshd stalled. Both deploy targets copy with =rsync --partial=, which
738resumes a stalled transfer.
739
740Under =-isolation podman=, the default and what bay1 runs, each build
741is confined to a container (see Container isolation below). Under
742=-isolation none= steps run directly on the host as the runner's user,
743so treat that machine as executing whatever your users push, and
744install the toolchains your builds need on it.
745
746A runner claims the oldest pending build among the repositories its key
747is attached to — for an admin key, the oldest in the instance. =-repos=
748narrows within that set, which is what makes a runner outside the server
749practical: one on a machine that should build a single project, or that
750holds credentials for one deployment, stays on it.
751
752Oldest-first is across everything the key may claim, so a repository
753with a deep queue holds every other repository the same runner serves;
754bay1 measured a 15-minute average wait on a day of merge request
755stacks from one repository. A runner attached to one repository cannot
756be starved. That is the rule, decided in krz/gitbay#207: a runner
757serving several repositories takes them oldest-first, and an operator
758who wants one repository never to wait on another runs a second
759runner attached to it alone. Nothing caps what an account queues,
760and nothing needs to (krz/gitbay#206): a schedule tick queues nothing
761while the job's last build is pending or running, so a repository
762with no runner holds one row per scheduled job rather than one per
763tick, and a build runs only on a runner its owner attaches, so a busy
764schedule spends the owner's compute. Pushes are bounded by what an
765account can push.
766
767#+begin_src sh
768gitbay-runner -remote git@gitbay.org -repos krz/site,krz/docs \
769 -workdir /var/lib/gitbay-runner/work
770#+end_src
771
772Add =-untrusted= only with =-isolation podman=.
773
774gitbay.org's runner is attached to the forge's own repositories and
775the isolation canary, nothing else, because it shares the host with
776the forge; its unit names no =-repos=, the attachments are the
777boundary. Any other repository builds on a runner its owner attaches.
778
779=-repos= narrows an admin runner; for a runner key the attachments are
780the boundary, held by the server, and =-repos= may only name
781repositories among them. =-untrusted= makes a runner claim merge
782request heads from forks; the bay1 unit sets it because it isolates in
783podman. A runner without it builds trusted commits only.
784
785=gitbay dashboard= and =ssh git@<host> admin runners= list every key
786that has polled as a runner: the account, the key's fingerprint, when it
787last polled, the repositories it may claim — its attachments for a runner
788key, the =-repos= it asked for or =any= for an admin key — and the build
789it holds; =admin runners remove <fingerprint>= (=forget= until the next release) drops the row for a key
790that polled by mistake, the key itself untouched. =admin runners= also
791heads the list with the queue: builds
792pending now, and over the last day how many were claimed, how long they
793waited to be claimed (average and worst), and
794how many the reaper ended instead of a runner reporting them. A build a runner claimed and never
795reported is failed by the scheduler's minute tick, whether or not any
796runner is still alive: within about two minutes of its log stream ending
797with no outcome reported — the runner reports right after closing the
798stream, retrying for half a minute if gitbayd is unreachable — or, if no
799stream was ever seen, at the deadline.
800
801Instance admin on the runner account only authorizes the claim/report
802protocol; it grants no repo access. A build that pushes back — a pages
803deploy, an archive publish, an automated MR branch — needs an explicit
804grant on that repo: =repo access grant <owner/name> ci write=. Private
805repos likewise need at least read for the clone.
806
807** Container isolation
808
809Builds run in a rootless podman container, one per job, with the
810workspace bind mounted and nothing else: the clone happens outside with
811the runner's key, so a step cannot read it. =-isolation none= keeps the
812old behaviour — steps on the host as the runner's user — for an instance
813where every repository is trusted. There is no automatic fallback: a
814runner started with =-isolation podman= that cannot find a working
815podman exits rather than running a build unsandboxed.
816
817gitbay's own jobs name =localhost/gitbay-ci:2=, built from
818=deploy/Containerfile.ci= on the runner host. A job's image must carry
819what its steps need: the suite drives real git, git-lfs, gpg and sshd and
820asserts they exist before running, so the stock runner default would fail
821it immediately. Build or rebuild it with:
822
823#+begin_src sh
824ssh -p 2222 root@<host> 'cat > /tmp/Containerfile.ci' < deploy/Containerfile.ci
825ssh -p 2222 root@<host> 'su - ci-runner -s /bin/sh -c \
826 "podman build -t localhost/gitbay-ci:2 -f /tmp/Containerfile.ci /tmp"'
827#+end_src
828
829The tag is deliberate rather than =:latest=: changing the file means
830bumping the tag in =.gitbay/ci.yml=, so a running branch's image does not
831change under it.
832
833=-image= names the image a job runs in when it declares none, and is
834required under =-isolation podman=: there is no built-in default,
835because an image this host does not have would fail every build. A job
836overrides it with =image:= in =.gitbay/ci.yml=, validated as a reference
837so a config file cannot turn it into podman arguments.
838
839=-cpus= and =-memory= cap one build (podman's units, e.g. =-cpus 2
840-memory 4g=); unset means uncapped. The runner applies them itself: it
841creates a cgroup per build under its own delegated service cgroup,
842writes the limits there, and starts every podman process for the build
843inside it, with podman's cgroup handling off. Podman's own =--memory=
844and =--cpus= never applied under rootless cgroupfs, which is what a
845system service gets (krz/gitbay#188). The unit therefore needs
846=Delegate=yes=, which the drop-in sets; without it the runner refuses
847to start when a limit is set, and logs that builds run unconfined when
848none is. bay1 runs =-cpus 3 -memory 6g= per build inside =MemoryMax=6G=
849and =CPUQuota=300%= on the unit, on a 7.7GB four-core host with no
850swap: the memory cap is what keeps the forge alive when a build
851allocates without bound, and it sits above the e2e suite's 5GB peak
852rather than at a fair share. =OOMPolicy=continue= keeps systemd from
853stopping the runner when a build is OOM-killed.
854
855A trusted build's home is its repository's, under
856=<workdir>/trusted-home/<owner>/<name>=, mounted into its containers as
857=HOME=: caches persist between trusted builds of one repository and are
858never read by another's. An untrusted build — a merge request head from
859a fork — gets =<workdir>/build-<id>-home=, new and empty, removed when
860the build ends. Homes under =<workdir>/home= are from runners before
861krz/gitbay#255, which shared them with untrusted builds; nothing reads
862them any more, and they can be deleted.
863
864*Images are provisioned, never pulled by a build.* The runner passes
865=--pull=never=. Two reasons, and the second is the better one: the
866service runs with =RestrictSUIDSGID=yes= so podman cannot unpack a layer
867holding a setuid file, which is nearly every distribution image; and on
868an instance where anyone can push a =ci.yml=, =image:= would otherwise
869mean "fetch and run anything from the internet". An operator pulls or
870builds what is allowed and a build picks among those. A job naming an
871image the host does not have fails with a message saying so.
872
873#+begin_src sh
874su - ci-runner -s /bin/sh -c "podman pull docker.io/library/alpine:3.20"
875su - ci-runner -s /bin/sh -c "podman images"
876#+end_src
877
878Prepare a host before pointing an isolating runner at it:
879
880#+begin_src sh
881ssh -p 2222 root@<host> 'sh -s' < deploy/runner-podman-setup.sh
882make deploy-runner
883#+end_src
884
885The script installs podman, delegates a subuid/subgid range to
886=ci-runner=, checks that user namespaces are enabled rather than
887assuming, enables lingering, and verifies rootless podman actually runs
888as that user. It is idempotent.
889
890It also installs nftables. =make deploy-runner= ships
891=deploy/gitbay-runner-egress.nft= to =/etc/gitbay-runner/egress.nft=
892with =gitbay-runner-egress.service=, which loads it and which the
893runner's unit requires; it checks the file with =nft -c=, reloads the
894unit, and runs =deploy/runner-egress-check.sh= as =ci-runner= before
895restarting the runner: =127.0.0.1:22= and the public 22 must answer,
8962222 must not. The table limits the runner's user to =127.0.0.1:22=,
897DNS on loopback, and 22, 80 and 443 on the host's public address; the
898Threat-Model page says why. A restart of =nftables.service= flushes it;
899=systemctl reload gitbay-runner-egress= restores it.
900
901The drop-in sets =NoNewPrivileges=no=, without which rootless podman
902cannot call =newuidmap= and the runner refuses to start. That is a
903considered trade, explained in the file and in the Threat-Model; if you
904run with =-isolation none=, set it back to =yes=.
905
906*Restarting the runner is safe.* On SIGTERM it stops claiming, finishes
907the build in flight, reports it, and exits; the drop-in's
908=TimeoutStopSec=50min= covers the longest build, and its =KillMode=mixed=
909is what makes the signal reach the runner alone — under systemd's default
910the build's container and the log session are signalled with it, and the
911runner drains a build that is already dead. So =make deploy-runner=
912waits for a running build rather than orphaning it, and a build's result
913is retried for half a minute if gitbayd is restarting at that moment. A
914second SIGTERM ends the runner at once, abandoning the build to the
915reaper. The suite checks all three: =TestRunnerDrainsOnSIGTERM= signals
916the process, =TestRunnerDropInLetsTheDrainHappen= reads the drop-in's
917=KillMode= and =TimeoutStopSec=, and =TestRunnerDrainUnderSystemd= runs
918the runner as a transient user unit under =systemd-run= and stops it
919under both kill modes. That last one needs a systemd user manager, so it
920skips in the container CI runs in; run it on a Linux host with
921=go test ./e2e -run TestRunnerDrainUnderSystemd -v=.
922
923*Validate podman mode on a scratch repository before pointing the runner
924at real ones.* Every deploy that switched the whole instance to
925containers and failed took CI down with it. Instead: create a throwaway
926repository the runner account can read (public, or granted read — a
927private one is "not found" to the runner and the build stays pending),
928give it one job that names the CI image, attach the runner's key to it,
929and deploy the runner with =-repos= naming only that repository. The
930production unit, with its real hardening, then claims nothing else;
931other repositories' builds queue until =-repos= is removed again, which
932is a pause, not an outage.
933
934#+begin_src sh
935gitbay repo create cmc/runner-scratch # then push a .gitbay/ci.yml naming the image
936# on the host:
937gitbay repo runner add cmc/runner-scratch < /var/lib/gitbay-runner/.ssh/id_ed25519.pub
938sed -i 's#^ExecStart=/usr/local/bin/gitbay-runner #&-repos cmc/runner-scratch #' /etc/systemd/system/gitbay-runner.service.d/override.conf
939systemctl daemon-reload && systemctl restart gitbay-runner
940gitbay build log cmc/runner-scratch 1 # green: remove -repos, redeploy, delete the scratch repository
941#+end_src
942
943*Do not deploy an isolating runner to a host that has not been
944prepared.* The runner is specified to refuse to start without a working
945podman rather than fall back to running builds unsandboxed — a fallback
946that silently drops isolation is worse than a stopped runner, because
947nothing surfaces it. On an unprepared host that refusal stops every
948build on the instance.
949
950The service drop-in carries =Delegate=yes= for rootless cgroup
951management and =ReadWritePaths= for podman's store under
952=/var/lib/gitbay-runner=, which =ProtectSystem=full= would otherwise
953make read-only. Those paths are prefixed =-= so they are ignored when
954absent: the drop-in installs on unprepared hosts too, and a unit that
955refused to start would stop every build.
956=gitbay-runner-prune.timer= prunes unused images weekly, as the runner's
957user: rootless storage belongs to that user, and root's prune would not
958see it. An unpruned image store on a 40GB host is a slow outage.
959
960* LFS storage
961
962Objects live content-addressed under =[lfs] root= (default
963=<server.root>/lfs=); =[lfs] max_object_bytes= caps a single object
964(512MB default). Storage sits behind a small interface — an
965S3-compatible backend is a drop-in with the server proxying, and
966presigned URLs a later optimization. LFS objects do not travel with
967push mirrors (mirrors move git refs only), and gc does not yet collect
968orphaned objects.
969
970* Pages
971
972=[pages] domain = "example.site"= serves public repos' =pages= branches
973on =<owner>.<domain>=. DNS needs a wildcard record =*.<domain>= to the
974server; ACME issues per-subdomain certificates on demand (only for
975owners that exist). The domain must not be the site host or a parent of
976it — pages content runs its own scripts and must stay off the forge's
977origin.
978
979Users with repo admin claim custom domains with =repo domain add=.
980Claims activate only after a DNS TXT challenge proves control of the
981domain (=repo domain verify=, audit-logged); pending claims serve
982nothing, get no certificates, and expire after 7 days. ACME issues
983certificates only for verified hosts, so stray DNS pointed at the
984server gets nothing.
985
986* Security
987
988The [[Threat-Model]] file is the reference for what the forge
989trusts and refuses to do. Operational checklist:
990
991- *Software checks.* =deploy/audit.sh= runs =go vet=, =govulncheck=
992 (the module list is deliberately short — review it on each release),
993 and a short fuzz pass over every attacker-facing parser (pkt-line,
994 commit, SSHSIG armor, OpenPGP key, SSH tokenizer). Run it before
995 tagging a release. CI's own =vuln= job runs =govulncheck= nightly
996 against main rather than per push, because =@latest= scans today's
997 advisory database and an advisory lands without anyone pushing;
998 =build trigger krz/gitbay vuln= runs it on demand.
999- *Web responses* carry a scripts-forbidden CSP, =X-Frame-Options:
1000 DENY=, =nosniff=, =no-referrer=, and HSTS when TLS is on — no
1001 configuration needed.
1002- *Host sandboxing.* The systemd unit in =deploy/cloud-init.yaml= runs
1003 gitbayd unprivileged with =ProtectSystem=strict=, =PrivateDevices=,
1004 =LockPersonality=, =MemoryDenyWriteExecute=,
1005 =SystemCallFilter=@system-service=, and =RestrictAddressFamilies= to
1006 INET/INET6/UNIX. It keeps =CAP_NET_BIND_SERVICE= only, to bind 22/80/443.
1007- *OS patches* apply via =unattended-upgrades= (security origins,
1008 auto-reboot 04:30 if required).
1009- *Admin sshd (2222)* is throttled by =MaxStartups=/=MaxAuthTries= and
1010 watched by =fail2ban=; gitbayd's own port 22 is throttled by
1011 =limits.ssh_auth_rate= (auth failures per IP per minute), and every
1012 account's writes by =limits.write_rate=.
1013- *Monitoring.* =gitbay-monitor.timer= writes a reading hourly to
1014journald and, when =/etc/gitbay/monitor.url= exists, posts it to that
1015webhook: disk, service, the daemon's own =/healthz= answer, certificate
1016expiry, and the age of the newest full backup and database snapshot.
1017It exits non-zero on an alert so the unit shows in =systemctl
1018--failed=: a stopped service, =/healthz= not answering =ok=, disk ≥ 85%,
1019a certificate under 21 days, a full backup over 25 hours old, or a
1020database snapshot over 2 hours old.
1021
1022=GET /healthz= is unauthenticated and cache-free: whether the database
1023answers and which commit serves, 503 when it does not.
1024- *Database.* =gitbay.db= and its WAL live under =/var/lib/gitbay= (mode
1025 0750, owned by =gitbay=). The nightly archive plus provider snapshots
1026 are the recovery path; for tighter RPO, add continuous replication
1027 (litestream) against the same file — it coexists with the WAL.
1028
1029* Odds and ends
1030
1031- deleting a fork marks MRs sourced from it =source_gone=; their diffs
1032 remain viewable and mergeable because the target repo owns the
1033 objects.
1034- =refs/merge-requests/*= is server-owned and unpushable by clients;
1035 only =admin mr prune= removes one (see Maintenance).
1036- audit-relevant activity (issue/MR lifecycle, imports, pushes) lands in
1037 the =events= table, which also feeds webhooks.
1038- the daemon idles under 10MB RSS; the smallest VPS tier is adequate.