1 | <?php |
||
2 | /** |
||
3 | * PingCheck.php |
||
4 | * |
||
5 | * Device up/down icmp check job |
||
6 | * |
||
7 | * This program is free software: you can redistribute it and/or modify |
||
8 | * it under the terms of the GNU General Public License as published by |
||
9 | * the Free Software Foundation, either version 3 of the License, or |
||
10 | * (at your option) any later version. |
||
11 | * |
||
12 | * This program is distributed in the hope that it will be useful, |
||
13 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
||
14 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.See the |
||
15 | * GNU General Public License for more details. |
||
16 | * |
||
17 | * You should have received a copy of the GNU General Public License |
||
18 | * along with this program. If not, see <https://www.gnu.org/licenses/>. |
||
19 | * |
||
20 | * @link https://www.librenms.org |
||
21 | * |
||
22 | * @copyright 2018 Tony Murray |
||
23 | * @author Tony Murray <[email protected]> |
||
24 | */ |
||
25 | |||
26 | namespace App\Jobs; |
||
27 | |||
28 | use App\Models\Device; |
||
29 | use Carbon\Carbon; |
||
30 | use Illuminate\Bus\Queueable; |
||
31 | use Illuminate\Contracts\Queue\ShouldQueue; |
||
32 | use Illuminate\Database\Eloquent\Builder; |
||
33 | use Illuminate\Foundation\Bus\Dispatchable; |
||
34 | use Illuminate\Queue\InteractsWithQueue; |
||
35 | use Illuminate\Queue\SerializesModels; |
||
36 | use Illuminate\Support\Collection; |
||
37 | use LibreNMS\Alert\AlertRules; |
||
38 | use LibreNMS\Config; |
||
39 | use LibreNMS\RRD\RrdDefinition; |
||
40 | use LibreNMS\Util\Debug; |
||
41 | use Log; |
||
42 | use Symfony\Component\Process\Process; |
||
43 | |||
44 | class PingCheck implements ShouldQueue |
||
45 | { |
||
46 | use Dispatchable, InteractsWithQueue, Queueable, SerializesModels; |
||
0 ignored issues
–
show
introduced
by
![]() |
|||
47 | |||
48 | private $command; |
||
49 | private $wait; |
||
50 | private $rrd_tags; |
||
51 | |||
52 | /** @var \Illuminate\Database\Eloquent\Collection List of devices keyed by hostname */ |
||
53 | private $devices; |
||
54 | /** @var array List of device group ids to check */ |
||
55 | private $groups = []; |
||
56 | |||
57 | // working data for loop |
||
58 | /** @var Collection */ |
||
59 | private $tiered; |
||
60 | /** @var Collection */ |
||
61 | private $current; |
||
62 | private $current_tier; |
||
63 | /** @var Collection */ |
||
64 | private $deferred; |
||
65 | |||
66 | /** |
||
67 | * Create a new job instance. |
||
68 | * |
||
69 | * @param array $groups List of distributed poller groups to check |
||
70 | */ |
||
71 | public function __construct($groups = []) |
||
72 | { |
||
73 | if (is_array($groups)) { |
||
0 ignored issues
–
show
|
|||
74 | $this->groups = $groups; |
||
75 | } |
||
76 | |||
77 | // define rrd tags |
||
78 | $rrd_step = Config::get('ping_rrd_step', Config::get('rrd.step', 300)); |
||
79 | $rrd_def = RrdDefinition::make()->addDataset('ping', 'GAUGE', 0, 65535, $rrd_step * 2); |
||
80 | $this->rrd_tags = ['rrd_def' => $rrd_def, 'rrd_step' => $rrd_step]; |
||
81 | |||
82 | // set up fping process |
||
83 | $timeout = Config::get('fping_options.timeout', 500); // must be smaller than period |
||
84 | $retries = Config::get('fping_options.retries', 2); // how many retries on failure |
||
85 | |||
86 | $this->command = ['fping', '-f', '-', '-e', '-t', $timeout, '-r', $retries]; |
||
87 | $this->wait = Config::get('rrd.step', 300) * 2; |
||
88 | } |
||
89 | |||
90 | /** |
||
91 | * Execute the job. |
||
92 | * |
||
93 | * @return void |
||
94 | */ |
||
95 | public function handle() |
||
96 | { |
||
97 | $ping_start = microtime(true); |
||
98 | |||
99 | $this->fetchDevices(); |
||
100 | |||
101 | $process = new Process($this->command, null, null, null, $this->wait); |
||
102 | |||
103 | d_echo($process->getCommandLine() . PHP_EOL); |
||
104 | |||
105 | // send hostnames to stdin to avoid overflowing cli length limits |
||
106 | $ordered_device_list = $this->tiered->get(1, collect())->keys()// root nodes before standalone nodes |
||
107 | ->merge($this->devices->keys()) |
||
108 | ->unique() |
||
109 | ->implode(PHP_EOL); |
||
110 | |||
111 | $process->setInput($ordered_device_list); |
||
112 | $process->start(); // start as early as possible |
||
113 | |||
114 | foreach ($process as $type => $line) { |
||
115 | d_echo($line); |
||
116 | |||
117 | if (Process::ERR === $type) { |
||
118 | // Check for devices we couldn't resolve dns for |
||
119 | if (preg_match('/^(?<hostname>[^\s]+): (?:Name or service not known|Temporary failure in name resolution)/', $line, $errored)) { |
||
120 | $this->recordData([ |
||
121 | 'hostname' => $errored['hostname'], |
||
122 | 'status' => 'unreachable', |
||
123 | ]); |
||
124 | } |
||
125 | continue; |
||
126 | } |
||
127 | |||
128 | if (preg_match( |
||
129 | '/^(?<hostname>[^\s]+) is (?<status>alive|unreachable)(?: \((?<rtt>[\d.]+) ms\))?/', |
||
130 | $line, |
||
131 | $captured |
||
132 | )) { |
||
133 | $this->recordData($captured); |
||
134 | |||
135 | $this->processTier(); |
||
136 | } |
||
137 | } |
||
138 | |||
139 | // check for any left over devices |
||
140 | if ($this->deferred->isNotEmpty()) { |
||
141 | d_echo("Leftover devices, this shouldn't happen: " . $this->deferred->flatten(1)->implode('hostname', ', ') . PHP_EOL); |
||
142 | d_echo('Devices left in tier: ' . collect($this->current)->implode('hostname', ', ') . PHP_EOL); |
||
143 | } |
||
144 | |||
145 | if (\App::runningInConsole()) { |
||
146 | printf("Pinged %s devices in %.2fs\n", $this->devices->count(), microtime(true) - $ping_start); |
||
147 | } |
||
148 | } |
||
149 | |||
150 | private function fetchDevices() |
||
151 | { |
||
152 | if (isset($this->devices)) { |
||
153 | return $this->devices; |
||
154 | } |
||
155 | |||
156 | /** @var Builder $query */ |
||
157 | $query = Device::canPing() |
||
158 | ->select(['devices.device_id', 'hostname', 'overwrite_ip', 'status', 'status_reason', 'last_ping', 'last_ping_timetaken', 'max_depth']) |
||
159 | ->orderBy('max_depth'); |
||
160 | |||
161 | if ($this->groups) { |
||
162 | $query->whereIn('poller_group', $this->groups); |
||
163 | } |
||
164 | |||
165 | $this->devices = $query->get()->keyBy(function ($device) { |
||
166 | return Device::pollerTarget(json_decode(json_encode($device), true)); |
||
167 | }); |
||
168 | |||
169 | // working collections |
||
170 | $this->tiered = $this->devices->groupBy('max_depth', true); |
||
171 | $this->deferred = collect(); |
||
172 | |||
173 | // start with tier 1 (the root nodes, 0 is standalone) |
||
174 | $this->current_tier = 1; |
||
175 | $this->current = $this->tiered->get($this->current_tier, collect()); |
||
176 | |||
177 | if (Debug::isVerbose()) { |
||
178 | $this->tiered->each(function (Collection $tier, $index) { |
||
179 | echo "Tier $index (" . $tier->count() . '): '; |
||
180 | echo $tier->implode('hostname', ', '); |
||
181 | echo PHP_EOL; |
||
182 | }); |
||
183 | } |
||
184 | |||
185 | return $this->devices; |
||
186 | } |
||
187 | |||
188 | /** |
||
189 | * Check if this tier is complete and move to the next tier |
||
190 | * If we moved to the next tier, check if we can report any of our deferred results |
||
191 | */ |
||
192 | private function processTier() |
||
193 | { |
||
194 | if ($this->current->isNotEmpty()) { |
||
195 | return; |
||
196 | } |
||
197 | |||
198 | $this->current_tier++; // next tier |
||
199 | |||
200 | if (! $this->tiered->has($this->current_tier)) { |
||
201 | // out of devices |
||
202 | return; |
||
203 | } |
||
204 | |||
205 | if (Debug::isVerbose()) { |
||
206 | echo "Out of devices at this tier, moving to tier $this->current_tier\n"; |
||
207 | } |
||
208 | |||
209 | $this->current = $this->tiered->get($this->current_tier); |
||
210 | |||
211 | // update and remove devices in the current tier |
||
212 | foreach ($this->deferred->pull($this->current_tier, []) as $data) { |
||
213 | $this->recordData($data); |
||
214 | } |
||
215 | |||
216 | // try to process the new tier in case we took care of all the devices |
||
217 | $this->processTier(); |
||
218 | } |
||
219 | |||
220 | /** |
||
221 | * If the device is on the current tier, record the data and remove it |
||
222 | * $data should have keys: hostname, status, and conditionally rtt |
||
223 | * |
||
224 | * @param array $data |
||
225 | */ |
||
226 | private function recordData(array $data) |
||
227 | { |
||
228 | if (Debug::isVerbose()) { |
||
229 | echo "Attempting to record data for {$data['hostname']}... "; |
||
230 | } |
||
231 | |||
232 | /** @var Device $device */ |
||
233 | $device = $this->devices->get($data['hostname']); |
||
234 | |||
235 | // process the data if this is a standalone device or in the current tier |
||
236 | if ($device->max_depth === 0 || $this->current->has($device->hostname)) { |
||
237 | if (Debug::isVerbose()) { |
||
238 | echo "Success\n"; |
||
239 | } |
||
240 | |||
241 | // mark up only if snmp is not down too |
||
242 | $device->status = ($data['status'] == 'alive' && $device->status_reason != 'snmp'); |
||
243 | $device->last_ping = Carbon::now(); |
||
244 | $device->last_ping_timetaken = isset($data['rtt']) ? $data['rtt'] : 0; |
||
245 | |||
246 | if ($device->isDirty('status')) { |
||
247 | // if changed, update reason |
||
248 | $device->status_reason = $device->status ? '' : 'icmp'; |
||
249 | $type = $device->status ? 'up' : 'down'; |
||
250 | Log::event('Device status changed to ' . ucfirst($type) . ' from icmp check.', $device->device_id, $type); |
||
251 | } |
||
252 | |||
253 | $device->save(); // only saves if needed (which is every time because of last_ping) |
||
254 | |||
255 | if (isset($type)) { // only run alert rules if status changed |
||
256 | echo "Device $device->hostname changed status to $type, running alerts\n"; |
||
257 | $rules = new AlertRules; |
||
258 | $rules->runRules($device->device_id); |
||
259 | } |
||
260 | |||
261 | // add data to rrd |
||
262 | app('Datastore')->put($device->toArray(), 'ping-perf', $this->rrd_tags, ['ping' => $device->last_ping_timetaken]); |
||
263 | |||
264 | // done with this device |
||
265 | $this->complete($device->hostname); |
||
266 | d_echo("Recorded data for $device->hostname (tier $device->max_depth)\n"); |
||
267 | } else { |
||
268 | if (Debug::isVerbose()) { |
||
269 | echo "Deferred\n"; |
||
270 | } |
||
271 | |||
272 | $this->defer($data); |
||
273 | } |
||
274 | } |
||
275 | |||
276 | /** |
||
277 | * Done processing $hostname, remove it from our active data |
||
278 | * |
||
279 | * @param string $hostname |
||
280 | */ |
||
281 | private function complete($hostname) |
||
282 | { |
||
283 | $this->current->offsetUnset($hostname); |
||
284 | $this->deferred->each->offsetUnset($hostname); |
||
285 | } |
||
286 | |||
287 | /** |
||
288 | * Defer this data processing until all parent devices are complete |
||
289 | * |
||
290 | * |
||
291 | * @param array $data |
||
292 | */ |
||
293 | private function defer(array $data) |
||
294 | { |
||
295 | $device = $this->devices->get($data['hostname']); |
||
296 | |||
297 | if ($this->deferred->has($device->max_depth)) { |
||
298 | // add this data to the proper tier, unless it already exists... |
||
299 | $tier = $this->deferred->get($device->max_depth); |
||
300 | if (! $tier->has($device->hostname)) { |
||
301 | $tier->put($device->hostname, $data); |
||
302 | } |
||
303 | } else { |
||
304 | // create a new tier containing this data |
||
305 | $this->deferred->put($device->max_depth, collect([$device->hostname => $data])); |
||
306 | } |
||
307 | } |
||
308 | } |
||
309 |