Replace synchronous log writing with zerolog's diode-based non-blocking
writer to prevent logging from blocking the main application during
log bursts. The diode writer buffers up to 1024 messages and logs a
warning when messages are dropped.
- Extract multi-writer logic into separate `multiWriter` function
- Wrap with `diode.NewWriter` for async buffering
- Update both `NewLogger` and `NewLoggerWithFixedLevel` to use diode
Streamlined the `loadRouteProviders()` function by:
- Replacing channel-based concurrency with a simpler sequential registration pattern after agent initialization
- Using `gperr.NewGroup` and `gperr.NewBuilder` for more idiomatic error handling
- Adding mutex protection for concurrent result building
- Removing the `storeProvider` helper method
Replaced hardcoded 10-second initialization timeout with a configurable `INIT_TIMEOUT` environment variable.
The new default is 1 minute, allowing operators to adjust startup behavior based on their infrastructure requirements.
Introduced a new session refresh mechanism in the Proxmox configuration to ensure the API session remains active. This includes:
- Added `SessionRefreshInterval` constant for configurable session refresh timing.
- Implemented `refreshSessionLoop` method to periodically refresh the session and handle errors with exponential backoff.
This enhancement improves the reliability of interactions with the Proxmox API by preventing session expiry.
Add new `/proxmox/tail` API endpoint for streaming file contents from Proxmox
nodes and LXC containers via WebSocket. Extend journalctl endpoint to support
filtering by multiple services simultaneously.
Changes:
- Add `GET /proxmox/tail` endpoint supporting node-level and LXC container file tailing
- Change `service` parameter from string to array in journalctl endpoints
- Add input validation (`checkValidInput`) to prevent command injection
- Refactor command formatting with proper shell quoting
Security: All command inputs are validated for dangerous characters before
- Add BaseURL field to Client for node-level route configuration
- Change VMID from int to *int to support three states:
- nil: auto-discover node or VM from hostname/IP/alias
- 0: node-level route (direct to Proxmox node API)
- >0: LXC/QEMU resource route with container control
- Change Service string to Services []string for multi-service support
- Implement proper node-level route handling: HTTPS scheme,
hostname from node BaseURL, default port 8006
- Move initial UpdateResources call to Init before starting loop
- Move proxmox auto-discovery earlier in route validation
BREAKING: NodeConfig.VMID is now a pointer type; NodeConfig.Service
renamed to Services (backward compatible via alias)
Refactored the journalctl API to accept `node`, `vmid`, and `service` parameters as query strings in addition to path parameters. Added a new route `/proxmox/journalctl` that accepts all parameters via query string while maintaining backward compatibility with existing path-parameter routes.
- Changed `JournalctlRequest` struct binding from URI-only to query+URI
- Simplified Swagger documentation by consolidating multiple route definitions
- Existing path-parameter routes remain functional for backward compatibility
Decouple the types package from the internal/proxmox package by defining
a standalone ProxmoxConfig struct. This reduces circular dependencies
and allows the types package to define its own configuration structures
without importing the proxmox package.
The route validation logic now converts between types.ProxmoxConfig and
proxmox.NodeConfig where needed for internal operations.
Add new `/proxmox/stats/{node}` API endpoint for retrieving Proxmox node
statistics in JSON format. The endpoint returns kernel version, CPU
usage/model, memory usage, rootfs usage, uptime, and load averages.
The existing `/proxmox/stats/{node}/{vmid}` endpoint has been corrected `VMStats` to return`text/plain` instead of `application/json`.
Both endpoints support WebSocket streaming for real-time stats updates
with a 1-second poll interval.
This change enables Proxmox node-level operations without requiring a specific
LXC container VMID.
**Features added:**
- New `/proxmox/journalctl/{node}` API endpoint for streaming node journalctl
- Route configuration support for Proxmox nodes (VMID = 0)
- `ReverseLookupNode` function for node discovery by hostname/IP/alias
- `NodeJournalctl` method for executing journalctl on nodes
**Behavior changes:**
- VMID parameter in journalctl endpoints is now optional
- Routes targeting nodes (without specific containers) are now valid
**Bug fixes:**
- Fixed error message variable reference in route validation
The LXCCommand method contained duplicate websocket handling logic for connecting to Proxmox's VNC terminal proxy. This refactoring extracts the common websocket connection, streaming, and cleanup logic into a new NodeCommand method on the Node type, allowing LXCCommand to simply format the pct command and delegate.
The go-proxmox submodule was also updated to access the NewNode constructor, which provides a cleaner API for creating node instances with the HTTP client.
- Moves ~100 lines of websocket handling from lxc_command.go to node.go
- Adds reusable NodeCommand method for executing commands via VNC websocket
- LXCCommand now simply calls NodeCommand with formatted command
- Maintains identical behavior and output streaming semantics
Added a function to close idle HTTP connections in the LXCCommand method. This addresses potential goroutine leaks caused by the go-proxmox library's TermWebSocket not closing underlying HTTP/2 connections. The websocket closer is now wrapped to ensure proper cleanup of transport connections when the command execution is finished.
Add start, stop, and restart endpoints for LXC containers via the Proxmox API:
- POST /api/v1/proxmox/lxc/:node/:vmid/start
- POST /api/v1/proxmox/lxc/:node/:vmid/stop
- POST /api/v1/proxmox/lxc/:node/:vmid/restart
Added new Proxmox journalctl endpoint `/journalctl/:node/:vmid` for viewing all
journalctl output without requiring a service name. Made the service parameter
optional across both endpoints.
Introduced configurable `limit` query parameter (1-1000, default 100) to both
proxmox journalctl and docker logs APIs, replacing hardcoded 100-line tail.
Added container status check in LXCCommand to prevent command execution on
stopped containers, returning a clear status message instead.
Refactored route validation to use pre-fetched IPs and improved References()
method for proxmox routes with better alias handling.
- Add VMResource wrapper type with cached IP addresses for efficient lookups
- Implement concurrent IP fetching during resource updates (limited concurrency)
- Add ReverseLookupResource for discovering VMs by IP, hostname, or alias
- Prioritize interfaces API over config for IP retrieval (offline container fallback)
- Enable routes to auto-discover Proxmox resources when no explicit config provided
- Fix configuration type from value to pointer slice for correct proxmox client retrievel
- Ensure Proxmox providers are initialized before route validation
Implement a new API endpoint to retrieve real-time statistics for Proxmox
LXC containers, similar to `docker stats` functionality.
Changes:
- Add `GET /api/v1/proxmox/stats/:node/:vmid` endpoint with HTTP and WebSocket support
- Implement resource polling loop to cache VM metadata every 3 seconds
- Create `LXCStats()` method with streaming (websocket) and single-shot modes
- Format output as: STATUS|CPU%|MEM USAGE/LIMIT|MEM%|NET I/O|BLOCK I/O
- Add `GetResource()` method for efficient VM resource lookup by kind and ID
- Fix task creation bug using correct client reference
Example response:
running|31.1%|9.6GiB/20GiB|48.87%|4.7GiB/3.3GiB|25GiB/36GiB
Centralize Proxmox node configuration by moving `ProxmoxConfig` from `internal/types/idlewatcher.go` to a new `NodeConfig` struct in `internal/proxmox/node.go`.
- Add `proxmox` field to route; allowing `proxy.app.proxmox` labels and corresponding route file config
- Added `service` optional field to NodeConfig for service identification
- Integrated Proxmox config directly into Route struct with proper validation
- Propagate Proxmox settings to Idlewatcher during route validation
- Updated swagger documentation to reflect schema changes
Add new /api/v1/proxmox/journalctl/:node/:vmid/:service endpoint that
streams real-time journalctl output from Proxmox LXC containers via
WebSocket connection. This enables live monitoring of container services
from the GoDoxy WebUI.
Implementation includes:
- New proxmox API handler with path parameter validation
- WebSocket upgrade for streaming output
- LXCCommand helper for executing commands over Proxmox VNC websocket
- LXCJournalctl wrapper for convenient journalctl -u service -f invocation
- Updated API documentation with proxmox integration
refactor(proxmox): support for PAM authentication
- Added support for username and password authentication alongside existing token-based authentication.
- Updated validation rules to require either token or username/password for authentication.
- Modified the Init function to handle session creation based on the selected authentication method.
- Increased timeout duration for context in the Init function.
Add the go-proxmox library as a Git submodule to enable Proxmox
integration for container/VM management.
Submodule: https://github.com/yusing/go-proxmox
- Removed `display_name`, `is_docker`, and `is_excluded` fields from the `RouteAggregate` struct and corresponding Swagger documentation.
- Updated references in the README and code to reflect the removal of these fields, ensuring consistency across the codebase.
- Updated route retrieval in the API and idle watcher to use GetIncludeExcluded, allowing for the inclusion of excluded routes.
- Simplified the route status aggregation logic by directly using GetIncludeExcluded for display name resolution.
- Removed redundant code that separately handled excluded routes, streamlining the route management process.
- Introduced a new GET endpoint `/docker/stats/:id` to fetch statistics for a specified container by its ID or route alias.
- Implemented the `Stats` function in the `dockerapi` package to handle the request and return container stats in both JSON and WebSocket formats.
- Added error handling for invalid requests and container not found scenarios.
Removes the embedded HTTP handler and WebSocket streaming capability from the
in-memory logger, leaving only the core io.Writer interface and event subscription
via Events(). Simplifies buffer management by eliminating position-based tracking
and using slices.Clone() for safe message passing to listeners.
- Removes HandlerFunc(), ServeHTTP(), wsInitial(), wsStreamLog() methods
- Removes logEntryRange struct and connChans map (no longer needed)
- Refactors buffer field from embedded to explicit buf with named mutexes
- Adds buffered channel (64) for event listeners to prevent blocking
- Improves concurrency with double-checked locking in truncation logic
- Updated NewHandler function to accept a requireAuth parameter for authentication control.
- Introduced a new local API server that allows unauthenticated access when LocalAPIHTTPAddr is set.
- Adjusted server startup logic to handle both authenticated and unauthenticated API routes.
Major refactoring of the access logging infrastructure to improve code organization and add proper console/stdout logging support.
- Renamed `Writer` interface to `File` and consolidated with `SupportRotate`
- Renamed `Log(req, res)` to `LogRequest(req, res)` for clarity
- Added new `ConsoleLogger` with zerolog console writer for formatted stdout output
- Moved type definitions to new `types.go` file
- Changed buffer handling from `[]byte` returns to `*bytes.Buffer` parameters
- Renamed internal files for clarity (`access_logger.go` → `file_access_logger.go`)
- Fixed fileserver access logging timing: moved logging after handler execution with defer
- Correct response handling in Fileserver
- Remove deprecated field `buffer_size`
- Simplify and removed unnecessary code
All callers have been updated to use the new APIs.
Replace per-scan byte slice allocations with a sized buffer pool,
significantly reducing memory pressure during log file scanning.
- Add Release() method to return buffers to pool (callers must invoke)
- Remove Reset() method - create new scanner instead for simpler lifecycle
- Refactor chunk prepending to reuse pooled buffers instead of append
Benchmark results show allocations dropped from ~26k to 1 per scan
for small chunk sizes, with better throughput.
BREAKING CHANGE: Reset() removed; callers must call Release() and
create a new BackScanner instance instead.
