mirror of
https://github.com/Vateron-Media/XC_VM.git
synced 2026-10-28 03:03:18 +01:00
Closed
opened 2026-09-25 12:04:24 +02:00 by thegame_1980
·
0 comments
No Branch/Tag Specified
main
fix/serveonly-noderoot-certbot
fix/serveonly-install-preserve-config
fix/reseller-ui-prefs-persist
gh-pages
feat/lb-dataplane-serveonly-policy
docs/lb-single-binary-parity-audit
feat/serveonly-noderoot-sudoers
fix/serveonly-keep-nginx-ports
feat/lb-serve-only-node-mode
feat/lb-p6-install-php-free
fix/lb-policy-section-server-id
feat/lb-p2-selfheal-gate
feat/lb-p1-policy-section
feat/lb-certbot-webroot
feat/lb-migration-cleanup-recovery
feat/lb-serve-only-provisioning-gate
feat/lb-node-crontab-filter
docs/lb-stop-php-crons
docs/lb-rtmp-gates-design
feat/lb-single-binary-roadmap
2.7.2
2.7.1
2.7.0
2.6.7
2.6.6-p6
2.6.6-p5
2.6.6-p4
2.6.6-p3
2.6.6-p2
2.6.6-p1
2.6.6
2.6.5
2.6.4
2.6.3
2.6.2
2.6.1
2.6.0
2.5.3
2.5.2
2.5.1
2.5.0
2.4.3
2.4.2
2.4.1
2.4.0
2.3.9
2.3.8
2.3.7
2.3.6
2.3.5
2.3.4
2.3.3
2.3.2
2.3.1
2.2.2
2.2.1
2.2.0
2.1.5
2.1.4
2.1.3
2.1.2
2.1.1
2.1.0
2.0.2
2.0.1
2.0.0
1.2.16
1.2.15
1.2.14
1.2.13
1.2.12
1.2.11
1.2.10
1.2.9
1.2.8
1.2.7
1.2.6
1.2.5
1.2.4
1.2.3
1.2.2
1.2.1
1.2.0
1.1.4
1.1.3
1.1.2
1.1.1
1.1.0
1.0.4
1.0.3
1.0.2
1.0.1
1.0.0
Labels
Clear labels
Post-release verification
Review & Deploy
awaiting triage
bug
composer
dependencies
documentation
done
duplicate
enhancement
github-actions
good first issue
help wanted
in progress
incident
invalid
needs info
on hold
planned
pull-request
question
rejected
security
status: backlog
status: in progress
status: post-release verification
status: review & deploy
tech-debt
under consideration
wontfix
It is necessary to check whether the problem remains after the update
Board status (auto-synced from Projects v2)
Not yet reviewed/prioritized
Something isn't working
Composer bump: re-commit a production-only vendor (composer install --no-dev) after merging
Dependency update (Dependabot)
Improvements or additions to documentation
Successfully completed and ready for closure.
This issue or pull request already exists
New feature or request
GitHub Actions version bump
Good for newcomers
Extra attention is needed
Actively being worked on right now.
Production incident (outage/degradation)
This doesn't seem right
More information is needed before work can start.
Work is paused, pending external action or decision.
Approved and scheduled for future work.
Mirrored from GitHub Pull Request
Further information is requested
Will not be implemented.
Security-sensitive issue or fix
Board: Backlog (synced from the XC_VM board)
Board: In Progress (synced from the XC_VM board)
Board: released, waiting for confirmation (synced from the XC_VM board)
Board: Review & Deploy, merged and not released yet (synced from the XC_VM board)
Refactoring / cleanup / technical debt
Idea or proposal awaiting triage and decision.
This will not be worked on
No labels
pull-request
Milestone
No items
No Milestone
Projects
Clear projects
No projects
Assignees
thegame_1980
Clear assignees
thegame_1980
No Assignees
Notifications
Due Date
No due date set.
Dependencies
No dependencies set.
Reference: Vateron-Media/XC_VM#116
Reference in New Issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
📋 Pull Request Information
Original PR: https://github.com/Vateron-Media/XC_VM/pull/204
Author: @obscuremind
Created: 9/25/2026
Status: ❌ Closed
Base:
main← Head:main📝 Commits (10+)
12257afdocs(design): add the MAIN ↔ LB API communication design spec44fda47Merge pull request #7 from obscuremind/claude/main-lb-api-communication-r0bdpp71108b2fix(logs): import every spooled activity and client-request row091660fMerge pull request #8 from obscuremind/claude/main-lb-api-communication-r0bdpp7464754fix(logs): update only existing lines and bound import batches by size8ef2440fix(stream): read connection rows for disallow_2nd_ip_con in Redis modea936d8ffix(stream): move the disallow_2nd_ip_con Redis decision into acceptedLineIP()7ee9e71fix(users): publish FPM worker pids so Redis reaping spares LB viewers6bc29e8fix(users): date a remote worker pid by its last write and tell idle pools from unknownf8f3ca3fix(watchdog): correct bandwidth totals, rps, producers and fatals📊 Changes
227 files changed (+14050 additions, -1048 deletions)
View changed files
📝
.gitignore(+1 -0)📝
Makefile(+57 -29)➕
docs/adr/0004-cluster-api.md(+126 -0)📝
docs/en/administration/server-diagnostics.md(+27 -1)📝
docs/en/builds/build_system.md(+56 -25)➕
docs/en/development/cluster-seams.md(+54 -0)📝
docs/en/development/streaming-subsystem.md(+45 -2)📝
docs/en/guides/cli-tools.md(+25 -2)📝
docs/en/guides/dev-workflow.md(+1 -1)📝
docs/en/guides/feature-flags.md(+1 -0)➕
docs/superpowers/specs/2026-09-21-main-lb-api-communication-design.md(+1131 -0)📝
lb_configs/nginx.conf(+4 -6)📝
mkdocs.yml(+2 -0)📝
src/Cli/Commands/CacheHandlerCommand.php(+5 -3)➕
src/Cli/Commands/ClusterInitCommand.php(+46 -0)📝
src/Cli/Commands/CreatedCommand.php(+6 -4)📝
src/Cli/Commands/DelayCommand.php(+5 -2)📝
src/Cli/Commands/FanoutSyncCommand.php(+12 -4)📝
src/Cli/Commands/LbInstallFlow.php(+32 -2)📝
src/Cli/Commands/MonitorCommand.php(+8 -7)...and 80 more files
📄 Description
This PR starts work on the MAIN ↔ LB API communication plan. Under the plan, LB nodes stop connecting to MAIN's MySQL and Redis, and every exchange moves to a signed, encrypted cluster API. Node tokens derive from the licence and rotate on an interval set in Settings.
docs/superpowers/specs/2026-09-21-main-lb-api-communication-design.md adds the design as supplied: revision 3 of the condensed plan plus 28 consistency corrections. It is the source of truth for the phases (0–11) and decisions D1–D21.
Phase 0 ("fix prerequisite defects, add legacy-backed seams and gates; no behaviour change") lands on this branch as separate commits, each with its own unit tests. It fixes five defects:
Log-import crons keep only one spooled row per node per minute (ActivityCronJob, LinesLogsCronJob).
In Redis mode, disallow_2nd_ip_con throws a ValueError in live.php (and a TypeError for HMAC identities).
servers.php_pids stores PHP-FPM master pids, so in Redis mode MAIN's cron:users kills live PHP-served viewers on LBs.
SystemInfo/watchdog telemetry: bandwidth totals, requests-per-second, the stream-producer count and two fatals that stop the heartbeat.
OPENSSL_EXTRA can differ between MAIN and LBs. The fix adds verification, provisioning, a repair command and a 10-minute dual-accept window.
The rest of Phase 0 follows on this branch: MAIN-only gates for crons, the LB build manifest, stop-gaps for proxy_api.php and the SSH credential file, and the legacy-backed seams (SignalDispatcher, NodeRpc/NodeActions, the stream seams, LogSink, lazy DB, audit hooks).
Type of change
Bug fix
New feature
Refactor / cleanup
Documentation
Build / CI / tooling
Checklist
Code follows the project conventions (see .github/instructions/php-conventions.instructions.md).
Added/updated tests for the change where applicable.
Documentation updated where applicable (docs/, README).
Verification per commit: php tests/phpunit.phar -c tests/phpunit.xml.dist (baseline 892 tests green) and make gates. PHPStan and phpcs were not run: this environment could not install the dev tools.
Phase 0 of the MAIN ↔ LB API communication plan (docs/superpowers/specs/2026-09-21-main-lb-api-communication-design.md, merged in https://github.com/obscuremind/XC_VM/pull/7): fix prerequisite defects, add legacy-backed seams and gates. Nothing changes behaviour except the defect fixes themselves.
Commits land here one Phase 0 item at a time. Each item starts with a failing unit test and gets two adversarial reviews before it is final.
In this PR so far
fix(logs): ActivityCronJob and LinesLogsCronJob kept only the first spooled row per node per minute and deleted the rest. They now import every row in batches of 1000, skip rows that don't decode, and claim the spool by renaming it so concurrent writers aren't lost.
Coming next on this branch
live.php: in Redis mode, disallow_2nd_ip_con throws a ValueError.
php_pids: it holds PHP-FPM master pids, so in Redis mode MAIN's cron:users kills live PHP-served LB viewers.
SystemInfo: bandwidth totals, rps, the stream-producer count, and two fatals that break the heartbeat.
OPENSSL_EXTRA: a mismatch between MAIN and LB. The fix adds a fingerprint check, provisioning, a repair command and a 10-minute dual-accept window.
MAIN-only gates for crons, the stale LB build manifest, and the proxy_api.php and SSH-credential stop-gaps.
Legacy-backed seams: SignalDispatcher, NodeRpc/NodeActions, StreamSource/StreamStateWriter/StreamRowMerge/StreamCacheBuilder, LogSink, lazy DB in the streaming bootstrap, and trace-only DB/Redis connect audit hooks.
Type of change
Bug fix
New feature
Refactor / cleanup
Documentation
Build / CI / tooling
Checklist
Code follows the project conventions (see .github/instructions/php-conventions.instructions.md).
Added/updated tests for the change where applicable.
Documentation updated where applicable (docs/, README).
Checked locally: php tests/phpunit.phar -c tests/phpunit.xml.dist passes (899 tests) and make gates passes. PHPStan comes from this PR's CI, because the dev tools can't be installed in this environment.
Merged
Phase 0: fix viewer kills, live.php fatal, log import, telemetry and OPENSSL_EXTRA drift
#9
obscuremind
merged 13 commits into
main
from
claude/main-lb-api-communication-r0bdpp
Conversation
@obscuremind
obscuremind
commented
6 hours ago
•
Summary
This PR fixes the five Phase 0 defects of the MAIN ↔ LB API communication plan (docs/superpowers/specs/2026-09-21-main-lb-api-communication-design.md). Each fix started from a failing test and went through two adversarial reviews.
fix(users), 2 commits. servers.php_pids held the 4 PHP-FPM master pids. In Redis mode, MAIN's cron:users never found a connection's worker pid in that list, so it dropped every live PHP-served viewer on an LB about a minute after connecting. That covers VOD, timeshift, and live TS when fanout is off. Changes:
The watchdog now publishes the worker pids (ProcessManager::phpFpmWorkerPIDs()).
A remote pid is dated by its last write.
An idle pool is no longer mistaken for an unknown one.
php_pids is kept out of the per-request servers cache and the admin API.
fix(stream), 2 commits. With disallow_2nd_ip_con on in Redis mode, live.php threw ValueError: Array sizes are inconsistent, and a TypeError for HMAC identities. ConnectionTracker::acceptedLineIP() now reads the connection rows behind the LINE# keys and accepts the oldest connection's IP, as MySQL mode already does.
fix(logs), follow-up to https://github.com/obscuremind/XC_VM/pull/8. The importer now updates only lines rows that exist, and it caps each batch by byte size as well as by row count.
fix(watchdog), 2 commits. Telemetry fixes:
The bandwidth totals are summed over the selected interfaces, and the received total reads rx_bytes.
requests_per_second now has a value, from a persistent CounterRateSampler.
total_running_streams counts ffmpeg processes and xc_fanout remux producers, read from /proc.
Two fatals that stopped the heartbeat are fixed.
The watchdog_data key set and order are unchanged.
OPENSSL_EXTRA drift between MAIN and LB, 5 commits. A MAIN installed with the current installer uses a random config/openssl_extra. LBs added with server:install fall back to the literal default. On such clusters, every token MAIN mints fails readToken on the LB. Changes:
feat(diagnose): each node publishes a one-way fingerprint of its value (server_hardware.openssl_extra_fp, never the value itself). server:diagnose reports a mismatch.
fix(install): LbInstallFlow::provisionOpensslExtra() gives new LBs MAIN's value.
fix(servers): a MAIN-only server:sync-openssl-extra repairs existing LBs through a root signal (set_openssl_extra). The command is stripped from the LB archive. The row is never handed to proxies by proxy_api.php. The command refuses to run when MAIN's FPM uses a different value from its CLI.
feat(stream): after a switch, Encryption::readToken also accepts the value the node replaced, for 10 minutes. It retries only after the current value fails. There is a documented ~1/256 miss for legacy (non-sealed) tokens.
Docs: server-diagnostics.md and cli-tools.md.
Type of change
Bug fix
New feature
Refactor / cleanup
Documentation
Build / CI / tooling
Checklist
Code follows the project conventions (see .github/instructions/php-conventions.instructions.md).
Added/updated tests for the change where applicable.
Documentation updated where applicable (docs/, README).
Validated locally on
6c391f25:PHPUnit: 995 tests pass (php tests/phpunit.phar -c tests/phpunit.xml.dist).
make gates passes; the LB manifest has 705 files.
PHPStan level 5 reports no errors.
phpcs (PSR-12 + Generic) passes on the changed files.
GitHub Actions runners are not starting jobs on this fork (see https://github.com/obscuremind/XC_VM/pull/8 and https://github.com/obscuremind/XC_VM/pull/9 comments), so these local runs stand in for CI for now.
🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.
[PR #204] Phase 0to [PR #204] [CLOSED] Phase 0[PR #204] [CLOSED] Phase 0to [PR #204] Phase 0[PR #204] Phase 0to [PR #204] [CLOSED] Phase 0[PR #204] [CLOSED] Phase 0to [PR #204] Phase 0[PR #204] Phase 0to [PR #204] [CLOSED] Phase 0