Merge pull request 'Removal of logs' (#318) from Chewing_Bever/vieter:logs-removal into dev

Reviewed-on: vieter-v/vieter#318

docs/api/source/includes/_jobs.md

@@ -0,0 +1,78 @@
+# Jobs
+
+<aside class="notice">
+
+All routes in this section require authentication.
+
+</aside>
+
+## Manually schedule a job
+
+```shell
+curl \
+  -H 'X-Api-Key: secret' \
+  https://example.com/api/v1/jobs/queue?target=10&force&arch=x86_64
+```
+
+Manually schedule a job on the server.
+
+### HTTP Request
+
+`POST /api/v1/jobs/queue`
+
+### Query Parameters
+
+Parameter | Description
+--------- | -----------
+target | Id of target to schedule build for
+arch | Architecture to build on
+force | Whether it's a forced build (true if present)
+
+## Poll for new jobs
+
+<aside class="warning">
+
+This endpoint is used by the agents and should not be used manually. It's just
+here for completeness. Requests to this endpoint modify the build queue,
+meaning manual requests can cause builds to be skipped.
+
+</aside>
+
+```shell
+curl \
+  -H 'X-Api-Key: secret' \
+  https://example.com/api/v1/jobs/poll?arch=x86_64&max=2
+```
+
+> JSON output format
+
+```json
+{
+  "message": "",
+  "data": [
+    {
+      "target_id": 1,
+      "kind": "git",
+      "url": "https://aur.archlinux.org/discord-ptb.git",
+      "branch": "master",
+      "path": "",
+      "repo": "bur",
+      "base_image": "archlinux:base-devel",
+      "force": true
+    }
+  ]
+}
+```
+
+Poll the server for new builds.
+
+### HTTP Request
+
+`GET /api/v1/jobs/poll`
+
+### Query Parameters
+
+Parameter | Description
+--------- | -----------
+arch | For which architecture to receive jobs
+max | How many jobs to receive at most

docs/api/source/includes/_logs.md

@@ -125,8 +125,8 @@ id | ID of requested log
 
 <aside class="warning">
 
-You should probably not use this endpoint, as it's used by the build system to
-publish its logs.
+This endpoint is used by the agents and should not be used manually unless you
+know what you're doing. It's just here for completeness.
 
 </aside>
 
@@ -149,3 +149,24 @@ target | id of target this build is for
 ### Request body
 
 Plaintext contents of the build log.
+
+## Remove a build log
+
+```shell
+curl \
+  -XDELETE \
+  -H 'X-Api-Key: secret' \
+  https://example.com/api/v1/logs/1
+```
+
+Remove a build log from the server.
+
+### HTTP Request
+
+`DELETE /api/v1/logs/:id`
+
+### URL Parameters
+
+Parameter | Description
+--------- | -----------
+id | id of log to remove

docs/api/source/includes/_targets.md

@@ -27,6 +27,7 @@ curl \
       "kind": "git",
       "url": "https://aur.archlinux.org/discord-ptb.git",
       "branch": "master",
+      "path" : "",
       "repo": "bur",
       "schedule": "",
       "arch": [
@@ -73,8 +74,9 @@ curl \
     "kind": "git",
     "url": "https://aur.archlinux.org/discord-ptb.git",
     "branch": "master",
+    "path": "",
     "repo": "bur",
-    "schedule": "0 3",
+    "schedule": "0 2",
     "arch": [
       {
         "id": 1,
@@ -124,6 +126,7 @@ Parameter | Description
 kind | Kind of target to add; one of 'git', 'url'.
 url | URL of the Git repository.
 branch | Branch of the Git repository.
+path | Subdirectory inside Git repository to use.
 repo | Vieter repository to publish built packages to.
 schedule | Cron build schedule (syntax explained [here](https://rustybever.be/docs/vieter/usage/builds/schedule/))
 arch | Comma-separated list of architectures to build package on.
@@ -149,12 +152,20 @@ Parameter | Description
 kind | Kind of target; one of 'git', 'url'.
 url | URL of the Git repository.
 branch | Branch of the Git repository.
+path | Subdirectory inside Git repository to use.
 repo | Vieter repository to publish built packages to.
 schedule | Cron build schedule
 arch | Comma-separated list of architectures to build package on.
 
 ## Remove a target
 
+```shell
+curl \
+  -XDELETE \
+  -H 'X-Api-Key: secret' \
+  https://example.com/api/v1/targets/1
+```
+
 Remove a target from the server.
 
 ### HTTP Request

docs/api/source/index.html.md

@@ -11,6 +11,7 @@ includes:
   - repository
   - targets
   - logs
+  - jobs
 
 search: true
 

docs/content/configuration.md

@@ -32,11 +32,11 @@ configuration variable required for each command.
 
 ### `vieter server`
 
+* `port`: HTTP port to run on
+    * Default: `8000`
 * `log_level`: log verbosity level. Value should be one of `FATAL`, `ERROR`,
   `WARN`, `INFO` or `DEBUG`.
     * Default: `WARN`
-* `log_file`: log file to write logs to.
-    * Default: `vieter.log` (in the current directory)
 * `pkg_dir`:  where Vieter should store the actual package archives.
 * `data_dir`: where Vieter stores the repositories, log file & database.
 * `api_key`: the API key to use when authenticating requests.
@@ -44,9 +44,27 @@ configuration variable required for each command.
     * Packages with architecture `any` are always added to this architecture.
       This prevents the server from being confused when an `any` package is
       published as the very first package for a repository.
-    * Git repositories added without an `arch` value use this value instead.
-* `port`: HTTP port to run on
-    * Default: `8000`
+    * Targets added without an `arch` value use this value instead.
+* `global_schedule`: build schedule for any target that does not have a
+  schedule defined. For information about this syntax, see
+  [here](/usage/builds/schedule).
+    * Default: `0 3` (3AM every night)
+* `base_image`: Docker image to use when building a package. Any Pacman-based
+  distro image should work, as long as `/etc/pacman.conf` is used &
+  `base-devel` exists in the repositories. Make sure that the image supports
+  the architecture of your cron daemon.
+    * Default: `archlinux:base-devel` (only works on `x86_64`). If you require
+      `aarch64` support, consider using
+      [`menci/archlinuxarm:base-devel`](https://hub.docker.com/r/menci/archlinuxarm)
+      ([GitHub](https://github.com/Menci/docker-archlinuxarm)). This is the
+      image used for the Vieter CI builds.
+* `max_log_age`: maximum age of logs (in days). Logs older than this will get
+  cleaned by the log removal daemon. If set to a negative value, no logs are
+  ever removed. The age of logs is determined by the time the build was
+  started.
+    * Default: `-1`
+* `log_removal_schedule`: cron schedule defining when to clean old logs.
+    * Default: `0 0` (every day at midnight)
 
 ### `vieter cron`
 

docs/content/installation.md

@@ -23,15 +23,15 @@ guarantees about stability, so beware!
 Thanks to the single-binary design of Vieter, this image can be used both for
 the repository server, the cron daemon and the agent.
 
-Below is an example compose file to set up both the repository server & the
-cron daemon:
+Below is a minimal compose file to set up both the repository server & a build
+agent:
 
 ```yaml
 version: '3'
 
 services:
   server:
-    image: 'chewingbever/vieter:dev'
+    image: 'chewingbever/vieter:0.5.0-rc.1'
     restart: 'always'
 
     environment:
@@ -41,18 +41,19 @@ services:
       - 'data:/data'
 
   cron:
-    image: 'chewingbever/vieter:dev'
+    image: 'chewingbever/vieter:0.5.0-rc.1'
     restart: 'always'
+    # Required to connect to the Docker daemon
     user: root
-    command: 'vieter cron'
+    command: 'vieter agent'
 
     environment:
       - 'VIETER_API_KEY=secret'
       # MUST be public URL of Vieter repository
       - 'VIETER_ADDRESS=https://example.com'
-      - 'VIETER_DEFAULT_ARCH=x86_64'
+      # Architecture for which the agent builds
+      - 'VIETER_ARCH=x86_64'
       - 'VIETER_MAX_CONCURRENT_BUILDS=2'
-      - 'VIETER_GLOBAL_SCHEDULE=0 3'
     volumes:
       - '/var/run/docker.sock:/var/run/docker.sock'
 
@@ -63,14 +64,17 @@ volumes:
 If you do not require the build system, the repository server can be used
 independently as well.
 
+Of course, Vieter allows a lot more configuration than this. This compose file
+is meant as a starting point for setting up your installation.
+
 {{< hint info >}}
 **Note**  
-Builds are executed on the cron daemon's system using the host's Docker daemon.
-A cron daemon on a specific architecture will only build packages for that
-specific architecture. Therefore, if you wish to build packages for both
-`x86_64` & `aarch64`, you'll have to deploy two cron daemons, one on each
-architecture. Afterwards, any Git repositories enabled for those two
-architectures will build on both.
+Builds are executed on the agent's system using the host's Docker daemon. An
+agent for a specific `arch` will only build packages for that specific
+architecture. Therefore, if you wish to build packages for both `x86_64` &
+`aarch64`, you'll have to deploy two agents, one on each architecture.
+Afterwards, any Git repositories enabled for those two architectures will build
+on both.
 {{< /hint >}}
 
 ## Binary
@@ -99,9 +103,9 @@ latest official release or `vieter-git` for the latest development release.
 ### AUR
 
 If you prefer building the packages locally (or on your own Vieter instance),
-there's the `[vieter](https://aur.archlinux.org/packages/vieter)` &
-`[vieter-git](https://aur.archlinux.org/packages/vieter-git)` packages on the
-AUR. These packages build using the `vlang-git` compiler package, so I can't
+there's the [`vieter`](https://aur.archlinux.org/packages/vieter) &
+[`vieter-git`](https://aur.archlinux.org/packages/vieter-git) packages on the
+AUR. These packages build using the `vlang` compiler package, so I can't
 guarantee that a compiler update won't temporarily break them.
 
 ## Building from source

docs/content/usage/builds/cleanup.md

@@ -0,0 +1,23 @@
+---
+weight: 20
+---
+
+# Cleanup
+
+Vieter stores the logs of every single package build. While this is great for
+debugging why builds fail, it also causes an active or long-running Vieter
+instance to accumulate thousands of logs.
+
+To combat this, a log removal daemon can be enabled that periodically removes
+old build logs. By starting your server with the `max_log_age` variable (see
+[Configuration](/configuration#vieter-server)), a daemon will get enabled that
+periodically removes logs older than this setting. By default, this will happen
+every day at midnight, but this behavior can be changed using the
+`log_removal_schedule` variable.
+
+{{< hint info >}}
+**Note**  
+The daemon will always run a removal of logs on startup. Therefore, it's
+possible the daemon will be *very* active when first enabling this setting.
+After the initial surge of logs to remove, it'll calm down again.
+{{< /hint >}}

docs/content/usage/builds/schedule.md

@@ -1,3 +1,7 @@
+---
+weight: 10
+---
+
 # Cron schedule syntax
 
 The Vieter cron daemon uses a subset of the cron expression syntax to schedule

src/client/logs.v

@@ -41,3 +41,8 @@ pub fn (c &Client) add_build_log(target_id int, start_time time.Time, end_time t
 
 	return data
 }
+
+// remove_build_log removes the build log with the given id from the server.
+pub fn (c &Client) remove_build_log(id int) ! {
+	c.send_request<string>(.delete, '/api/v1/logs/$id', {})!
+}

src/console/logs/logs.v

@@ -138,6 +138,18 @@ pub fn cmd() cli.Command {
 					list(conf, filter, raw)!
 				}
 			},
+			cli.Command{
+				name: 'remove'
+				required_args: 1
+				usage: 'id'
+				description: 'Remove a build log that matches the given id.'
+				execute: fn (cmd cli.Command) ! {
+					config_file := cmd.flags.get_string('config-file')!
+					conf := vconf.load<Config>(prefix: 'VIETER_', default_path: config_file)!
+
+					remove(conf, cmd.args[0])!
+				}
+			},
 			cli.Command{
 				name: 'info'
 				required_args: 1
@@ -204,3 +216,9 @@ fn content(conf Config, id int) ! {
 
 	println(content)
 }
+
+// remove removes a build log from the server's list.
+fn remove(conf Config, id string) ! {
+	c := client.new(conf.address, conf.api_key)
+	c.remove_build_log(id.int())!
+}

src/models/logs.v

@@ -1,6 +1,7 @@
 module models
 
 import time
+import os
 
 pub struct BuildLog {
 pub mut:
@@ -28,6 +29,13 @@ pub fn (bl &BuildLog) str() string {
 	return str
 }
 
+// path returns the path to the log file, relative to the logs directory
+pub fn (bl &BuildLog) path() string {
+	filename := bl.start_time.custom_format('YYYY-MM-DD_HH-mm-ss')
+
+	return os.join_path(bl.target_id.str(), bl.arch, filename)
+}
+
 [params]
 pub struct BuildLogFilter {
 pub mut:

src/server/api_logs.v

@@ -86,7 +86,7 @@ fn (mut app App) v1_post_log() web.Result {
 	}
 
 	// Store log in db
-	log := BuildLog{
+	mut log := BuildLog{
 		target_id: target_id
 		start_time: start_time
 		end_time: end_time
@@ -95,25 +95,20 @@ fn (mut app App) v1_post_log() web.Result {
 	}
 
 	// id of newly created log
-	log_id := app.db.add_build_log(log)
-
-	repo_logs_dir := os.join_path(app.conf.data_dir, logs_dir_name, target_id.str(), arch)
+	log.id = app.db.add_build_log(log)
+	log_file_path := os.join_path(app.conf.data_dir, logs_dir_name, log.path())
 
 	// Create the logs directory of it doesn't exist
-	if !os.exists(repo_logs_dir) {
-		os.mkdir_all(repo_logs_dir) or {
-			app.lerror("Couldn't create dir '$repo_logs_dir'.")
+	if !os.exists(os.dir(log_file_path)) {
+		os.mkdir_all(os.dir(log_file_path)) or {
+			app.lerror('Error while creating log file: $err.msg()')
 
 			return app.status(.internal_server_error)
 		}
 	}
 
-	// Stream log contents to correct file
-	file_name := start_time.custom_format('YYYY-MM-DD_HH-mm-ss')
-	full_path := os.join_path_single(repo_logs_dir, file_name)
-
 	if length := app.req.header.get(.content_length) {
-		util.reader_to_file(mut app.reader, length.int(), full_path) or {
+		util.reader_to_file(mut app.reader, length.int(), log_file_path) or {
 			app.lerror('An error occured while receiving logs: $err.msg()')
 
 			return app.status(.internal_server_error)
@@ -122,5 +117,22 @@ fn (mut app App) v1_post_log() web.Result {
 		return app.status(.length_required)
 	}
 
-	return app.json(.ok, new_data_response(log_id))
+	return app.json(.ok, new_data_response(log.id))
+}
+
+// v1_delete_log allows removing a build log from the system.
+['/api/v1/logs/:id'; auth; delete]
+fn (mut app App) v1_delete_log(id int) web.Result {
+	log := app.db.get_build_log(id) or { return app.status(.not_found) }
+	full_path := os.join_path(app.conf.data_dir, logs_dir_name, log.path())
+
+	os.rm(full_path) or {
+		app.lerror('Failed to remove log file $full_path: $err.msg()')
+
+		return app.status(.internal_server_error)
+	}
+
+	app.db.delete_build_log(id)
+
+	return app.status(.ok)
 }

src/server/cli.v

@@ -5,14 +5,16 @@ import conf as vconf
 
 struct Config {
 pub:
-	log_level       string = 'WARN'
-	pkg_dir         string
-	data_dir        string
-	api_key         string
-	default_arch    string
-	global_schedule string = '0 3'
-	port            int    = 8000
-	base_image      string = 'archlinux:base-devel'
+	port                 int    = 8000
+	log_level            string = 'WARN'
+	pkg_dir              string
+	data_dir             string
+	api_key              string
+	default_arch         string
+	global_schedule      string = '0 3'
+	base_image           string = 'archlinux:base-devel'
+	max_log_age          int    = -1
+	log_removal_schedule string = '0 0'
 }
 
 // cmd returns the cli submodule that handles starting the server

src/server/log_removal.v

@@ -0,0 +1,62 @@
+module server
+
+import time
+import models { BuildLog }
+import os
+import cron.expression { CronExpression }
+
+const fallback_log_removal_frequency = 24 * time.hour
+
+// log_removal_daemon removes old build logs every `log_removal_frequency`.
+fn (mut app App) log_removal_daemon(schedule CronExpression) {
+	mut start_time := time.Time{}
+
+	for {
+		start_time = time.now()
+
+		mut too_old_timestamp := time.now().add_days(-app.conf.max_log_age)
+
+		app.linfo('Cleaning logs before $too_old_timestamp')
+
+		mut offset := u64(0)
+		mut logs := []BuildLog{}
+		mut counter := 0
+		mut failed := 0
+
+		// Remove old logs
+		for {
+			logs = app.db.get_build_logs(before: too_old_timestamp, offset: offset, limit: 50)
+
+			for log in logs {
+				log_file_path := os.join_path(app.conf.data_dir, logs_dir_name, log.path())
+
+				os.rm(log_file_path) or {
+					app.lerror('Failed to remove log file $log_file_path: $err.msg()')
+					failed += 1
+
+					continue
+				}
+				app.db.delete_build_log(log.id)
+
+				counter += 1
+			}
+
+			if logs.len < 50 {
+				break
+			}
+
+			offset += 50
+		}
+
+		app.linfo('Cleaned $counter logs ($failed failed)')
+
+		// Sleep until the next cycle
+		next_time := schedule.next_from_now() or {
+			app.lerror("Log removal daemon couldn't calculate next time: $err.msg(); fallback to $server.fallback_log_removal_frequency")
+
+			start_time.add(server.fallback_log_removal_frequency)
+		}
+
+		time.sleep(next_time - time.now())
+	}
+}

src/server/server.v

@@ -55,6 +55,10 @@ pub fn server(conf Config) ! {
 		util.exit_with_message(1, 'Invalid global cron expression: $err.msg()')
 	}
 
+	log_removal_ce := expression.parse_expression(conf.log_removal_schedule) or {
+		util.exit_with_message(1, 'Invalid log removal cron expression: $err.msg()')
+	}
+
 	// Configure logger
 	log_level := log.level_from_tag(conf.log_level) or {
 		util.exit_with_message(1, 'Invalid log level. The allowed values are FATAL, ERROR, WARN, INFO & DEBUG.')
@@ -108,5 +112,9 @@ pub fn server(conf Config) ! {
 		util.exit_with_message(1, 'Failed to inialize job queue: $err.msg()')
 	}
 
+	if conf.max_log_age > 0 {
+		go app.log_removal_daemon(log_removal_ce)
+	}
+
 	web.run(app, conf.port)
 }

vieter.toml

@@ -12,3 +12,4 @@ address = "http://localhost:8000"
 api_update_frequency = 2
 image_rebuild_frequency = 1
 max_concurrent_builds = 3
+max_log_age = 64