When multiple scheduled tasks or database backups run concurrently on
the same server, they compete for the same SSH multiplexed connection
socket, causing race conditions and SSH exit code 255 errors.
This fix adds a `disableMultiplexing` parameter to bypass SSH
multiplexing for jobs that may run concurrently:
- Add `disableMultiplexing` param to `generateSshCommand()`
- Add `disableMultiplexing` param to `instant_remote_process()`
- Update `ScheduledTaskJob` to use `disableMultiplexing: true`
- Update `DatabaseBackupJob` to use `disableMultiplexing: true`
- Add debug logging to track execution without multiplexing
- Add unit tests for the new parameter
Each backup and scheduled task now gets an isolated SSH connection,
preventing contention on the shared multiplexed socket.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Added a new `collapsible` property to GetLogs component that allows disabling the expandable header, useful for log viewers in dedicated pages and slide-overs. Applied this to Sentinel logs, Proxy logs, and Coolify Proxy log pages. Also improved the toolbar by moving the lines counter to the left side with an inline prefix label and repositioning the match counter next to it for better organization.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Replace text "Documentation" links with SVG icons positioned in top-right corner, matching the established design pattern used in the service list. Update all four PostgreSQL type options: PostgreSQL 17, Supabase, PostGIS, and PGVector.
🤖 Generated with Claude Code
Co-Authored-By: Claude <noreply@anthropic.com>
Filter out null and empty environment variables when generating Nixpacks build
configuration to prevent JSON parsing errors. Environment variables with null or
empty values were being passed as `--env KEY=` which created invalid JSON with
null values, causing deployment failures.
This fix ensures only valid non-empty environment variables are included in both
user-defined and auto-generated COOLIFY_* environment variables.
🤖 Generated with Claude Code
Co-Authored-By: Claude <noreply@anthropic.com>
Updated the timestamp rendering condition to check both $timestamp and $showTimeStamps, ensuring the "Toggle Timestamps" button properly controls timestamp visibility in the log viewer.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Added maximum iteration limit (maxIterations = 3) to the decodeHtml function to prevent potential DoS attacks from deeply nested HTML entities. This matches the implementation in deployment/show.blade.php and ensures the function cannot be exploited for excessive CPU usage.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Update the deployment status to IN_PROGRESS immediately when a build is queued, rather than waiting for the job to start. This ensures the UI reflects the correct status without delay.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Moved .log-highlight styles from Livewire component views to resources/css/app.css for better separation of concerns and reusability. This follows Laravel and Livewire best practices by keeping styles in the appropriate location rather than inline in component views.
Changes:
- Added .log-highlight styles to resources/css/app.css
- Removed inline <style> tags from deployment/show.blade.php
- Removed inline <style> tags from get-logs.blade.php
- Added XSS security test for log viewer
- Applied code formatting with Laravel Pint
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Replace SVG logo with official PNG icon from RustFS GitHub organization.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Add RustFS service definition with Docker Compose configuration and SVG logo for Coolify's service marketplace. Includes S3-compatible object storage setup with health checks and configurable environment variables.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Features:
- Add client-side search filtering for runtime and deployment logs
- Add log download functionality (respects search filters)
- Make runtime log sections collapsible by default
- Auto-expand single container and lazy load logs on first expand
- Match deployment and runtime log view heights (40rem)
- Add debug toggle for deployment logs
- Improve scroll behavior with follow logs feature
🤖 Generated with Claude Code
Co-Authored-By: Claude <noreply@anthropic.com>
The guard was setting and immediately resetting the flag in the same
synchronous execution, providing no actual protection. Now the flag
stays true until proxy reaches a stable state (running/exited/error)
via WebSocket notification, with additional client-side guard.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
When stopping a service that's currently deploying, mark any IN_PROGRESS or QUEUED activities as CANCELLED. This prevents the status from remaining stuck at "starting" after containers are stopped.
Follows the existing pattern used in forceDeploy().
Extended the maximum allowed timeout for scheduled tasks from 3600 to 36000 seconds (10 hours). Also passes the configured timeout to instant_remote_process() so the SSH command respects the timeout setting.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Allows user-configured backup timeouts > 3600 to be respected. Previously, the SSH process used a hardcoded 3600 second timeout regardless of the job timeout setting. Now the timeout is passed through to instant_remote_process() for all backup operations.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
When restarting the proxy on localhost (server id 0), shows a warning
banner in the logs sidebar explaining that the connection may be
temporarily lost and to refresh the browser if logs stop updating.
Also cleans up notification noise by commenting out intermediate
status notifications (restarting, starting, stopping) that were
redundant with the visual status indicators.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
The error "container name already in use" occurred because the container
wasn't fully removed before docker compose up tried to create a new one.
Changes:
- Removed redundant stop/remove logic from START PHASE (was duplicating STOP PHASE)
- Made STOP PHASE more robust:
- Increased wait iterations from 10 to 15
- Added force remove on each iteration in case container got stuck
- Added final verification and force cleanup after the loop
- Added better logging to show removal progress
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Instead of calling StopProxy::run() (synchronous) then StartProxy::run()
(async), now we build a single command sequence that includes both stop
and start phases. This creates one Activity immediately via remote_process(),
so the UI receives the activity ID right away and can show logs in real-time
from the very beginning of the restart operation.
Key changes:
- Removed dependency on StopProxy and StartProxy actions
- Build combined command sequence inline in buildRestartCommands()
- Use remote_process() directly which returns Activity immediately
- Increased timeout from 60s to 120s to accommodate full restart
- Activity ID dispatched to UI within milliseconds of job starting
Flow is now:
1. Job starts → sets "restarting" status
2. Commands built synchronously (fast, no SSH)
3. remote_process() creates Activity and dispatches CoolifyTask job
4. Activity ID sent to UI immediately via WebSocket
5. UI opens activity monitor with real-time streaming logs
6. Logs show "Stopping proxy..." then "Starting proxy..." as they happen
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Set proxy status to 'restarting' and dispatch ProxyStatusChangedUI event
at the very beginning of handle() method, before StopProxy runs. This
notifies the UI immediately so users know a restart is in progress,
rather than waiting until after the stop operation completes.
Also simplified unit tests to focus on testable job configuration
(middleware, tries, timeout) without complex SchemalessAttributes mocking.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Add restartInitiated flag to prevent duplicate "Proxy restart initiated" messages
- Restore ProxyStatusChangedUI dispatch with activityId in RestartProxyJob
- This allows the UI to open the activity monitor and show logs during restart
- Simplified restart message (removed redundant "Monitor progress" text)
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
The Application::loadComposeFile method's finally block always saves
the model, which was persisting invalid base_directory values when
validation failed.
Changes:
- Add restoreBaseDirectory and restoreDockerComposeLocation parameters
to loadComposeFile() in both Application model and General component
- The finally block now restores BOTH base_directory and
docker_compose_location to the provided original values before saving
- When called from submit(), pass the original DB values so they are
restored on failure instead of the new invalid values
This ensures invalid paths are never persisted to the database.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
When restarting the proxy on localhost (where Coolify is running), the UI becomes inaccessible because the connection is lost. This change makes all proxy restarts run as background jobs with WebSocket notifications, allowing the operation to complete even after connection loss.
Changes:
- Enhanced ProxyStatusChangedUI event to carry activityId for log monitoring
- Updated RestartProxyJob to dispatch status events and track activity
- Simplified Navbar restart() to always dispatch job for all servers
- Enhanced showNotification() to handle activity monitoring and new statuses
- Added comprehensive unit and feature tests
Benefits:
- Prevents localhost lockout during proxy restarts
- Consistent behavior across all server types
- Non-blocking UI with real-time progress updates
- Automatic activity log monitoring
- Proper error handling and recovery
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Move compose file validation BEFORE database save to prevent invalid
base directory and docker compose location values from being persisted
when validation fails.
Changes:
- Move compose file validation before $this->application->save()
- Restore original values when validation fails
- Add resetErrorBag() to clear stale validation errors
This fixes two bugs:
1. Invalid paths were saved to DB even when validation failed
2. Error messages persisted after correcting to valid path
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Apply the same frontend path normalization pattern from commit f6398f7cf
to the General Settings page for consistency across all forms.
Changes:
- Add Alpine.js path normalization to Docker Compose section (base directory + compose location)
- Add Alpine.js path normalization to non-Docker Compose section (base directory + dockerfile location)
- Change wire:model to wire:model.defer to prevent backend requests during tab navigation
- Add @blur event handlers for immediate path normalization feedback
- Backend normalization remains as defensive fallback
This ensures consistent validation behavior and fixes potential tab focus
issues on the General Settings page.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Change wire:model.blur to wire:model.defer to prevent backend requests
during form navigation. Add Alpine.js path normalization functions that
run on blur, fixing tab focus issues while keeping path validation
purely on the frontend.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Move Carbon::setTestNow() to the beginning of each test before creating
test data. Previously, tests created servers using now() (real current
time) and only afterwards called Carbon::setTestNow(), making
sentinel_updated_at inconsistent with the test clock.
This caused staleness calculations to use different timelines:
- sentinel_updated_at was based on real time (e.g., Dec 2024)
- Test execution time was frozen at 2025-01-15
Now all timestamps use the same frozen test time, making staleness
checks predictable and tests reliable regardless of when they run.
Affected tests (all 7 test cases in the file):
- does not dispatch storage check when sentinel is in sync
- dispatches storage check when sentinel is out of sync
- dispatches storage check when sentinel is disabled
- respects custom hourly storage check frequency when sentinel is out of sync
- handles VALID_CRON_STRINGS mapping correctly when sentinel is out of sync
- respects server timezone for storage checks when sentinel is out of sync
- does not dispatch storage check outside schedule
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
When Sentinel is enabled and in sync, ServerStorageCheckJob was being
dispatched from two locations causing unnecessary duplication:
1. PushServerUpdateJob (every ~30s with real-time filesystem data)
2. ServerManagerJob (scheduled cron check via SSH)
This commit modifies ServerManagerJob to only dispatch ServerStorageCheckJob
when Sentinel is out of sync or disabled. When Sentinel is active and in sync,
PushServerUpdateJob provides real-time storage data, making the scheduled SSH
check redundant.
Benefits:
- Eliminates duplicate storage checks when Sentinel is working
- Reduces unnecessary SSH overhead
- Storage checks still run as fallback when Sentinel fails
- Maintains scheduled checks for servers without Sentinel
Updated tests to reflect new behavior:
- Storage check NOT dispatched when Sentinel is in sync
- Storage check dispatched when Sentinel is out of sync or disabled
- All timezone and frequency tests updated accordingly
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
