1 /* Copyright (c) 2007-2023. The SimGrid Team. All rights reserved. */
3 /* This program is free software; you can redistribute it and/or modify it
4 * under the terms of the license (GNU LGPL) which comes with this package. */
6 /* This is a version of the masterworkers that (hopefully) survives to the chaos monkey.
7 * It tests synchronous send/receive as well as synchronous computations.
9 * It is not written to be pleasant to read, but instead to resist the aggressions of the monkey:
10 * - Workers keep going until after a global variable `todo` reaches 0.
11 * - The master is a daemon that just sends infinitely tasks
12 * (SimGrid simulations stop as soon as all non-daemon actors are done).
13 * - The platform is created programmatically to remove path issues and control the problem size.
15 * Command-line configuration items:
16 * - host-count: how many actors to start (including the master
17 * - task-count: initial value of the `todo` global
18 * - deadline: time at which the simulation is known to be failed (to detect infinite loops).
20 * See the simgrid-monkey script for more information.
23 #include <simgrid/s4u.hpp>
24 #include <xbt/config.hpp>
25 #include <xbt/string.hpp>
27 namespace sg4 = simgrid::s4u;
29 XBT_LOG_NEW_DEFAULT_CATEGORY(s4u_test, "Messages specific for this s4u example");
31 static simgrid::config::Flag<int> cfg_host_count{"host-count", "Host count (master on one, workers on the others)", 3};
32 static simgrid::config::Flag<double> cfg_deadline{"deadline", "When to fail the simulation (infinite loop detection)",
34 static simgrid::config::Flag<int> cfg_task_count{"task-count", "Amount of tasks that must be executed to succeed", 1};
36 XBT_ATTRIB_NORETURN static void master(sg4::Mailbox* mailbox)
38 double comp_size = 1e6;
40 bool rebooting = sg4::Actor::self()->get_restart_count() > 0;
42 XBT_INFO("Master %s", rebooting ? "rebooting" : "booting");
43 if (not rebooting) // Starting for the first time
44 sg4::this_actor::on_exit(
45 [](bool forcefully) { XBT_INFO("Master dying %s.", forcefully ? "forcefully" : "peacefully"); });
47 while (true) { // This is a daemon
48 xbt_assert(sg4::Engine::get_clock() < cfg_deadline,
49 "Failed to run all tasks in less than %d seconds. Is this an infinite loop?", (int)cfg_deadline);
51 auto payload = std::make_unique<double>(comp_size);
53 XBT_INFO("Try to send a message");
54 mailbox->put(payload.get(), comm_size, 10.0);
56 } catch (const simgrid::TimeoutException&) {
57 XBT_INFO("Timeouted while sending a task");
58 } catch (const simgrid::NetworkFailureException&) {
59 XBT_INFO("Got a NetworkFailureException. Wait a second before starting again.");
60 sg4::this_actor::sleep_for(1);
66 static void worker(int id, sg4::Mailbox* mailbox, int& todo)
68 bool rebooting = sg4::Actor::self()->get_restart_count() > 0;
70 XBT_INFO("Worker %s", rebooting ? "rebooting" : "booting");
71 if (not rebooting) // Starting for the first time
72 sg4::this_actor::on_exit(
73 [id](bool forcefully) { XBT_INFO("worker %d dying %s.", id, forcefully ? "forcefully" : "peacefully"); });
76 xbt_assert(sg4::Engine::get_clock() < cfg_deadline,
77 "Failed to run all tasks in less than %d seconds. Is this an infinite loop?", (int)cfg_deadline);
79 XBT_INFO("Waiting a message on %s", mailbox->get_cname());
80 auto payload = mailbox->get_unique<double>(10);
81 xbt_assert(payload != nullptr, "mailbox->get() failed");
82 double comp_size = *payload;
84 XBT_INFO("Start execution...");
85 sg4::this_actor::execute(comp_size);
86 XBT_INFO("Execution complete.");
88 } catch (const simgrid::TimeoutException&) {
89 XBT_INFO("Timeouted while getting a task.");
90 } catch (const simgrid::NetworkFailureException&) {
91 XBT_INFO("Got a NetworkFailureException. Wait a second before starting again.");
92 sg4::this_actor::sleep_for(1);
97 int main(int argc, char* argv[])
99 sg4::Engine e(&argc, argv);
101 auto* rootzone = sg4::create_full_zone("root");
102 std::vector<sg4::Host*> worker_hosts;
104 int todo = cfg_task_count; // remaining amount of tasks to execute, a shared variable
105 xbt_assert(todo > 0, "Please give more than %d tasks to run", todo);
107 xbt_assert(cfg_host_count > 2, "You need at least 2 workers (i.e., 3 hosts) or the master will be auto-killed when "
108 "the only worker gets killed.");
110 // get it once for all, to reduce the amount of simcalls during actor reboot
111 sg4::Mailbox* mailbox = sg4::Mailbox::by_name("mailbox");
113 sg4::Host* master_host = rootzone->create_host("lilibeth 0", 1e9); // Host where the master will stay
114 for (int i = 1; i < cfg_host_count; i++) {
115 auto hostname = "lilibeth " + std::to_string(i);
116 auto* host = rootzone->create_host(hostname, 1e9);
117 auto* link = rootzone->create_link(hostname, "1MBps")->set_latency("24us")->seal();
118 rootzone->add_route(master_host, host, {link});
119 worker_hosts.push_back(host);
123 sg4::Actor::create("master", master_host, master, mailbox)->daemonize()->set_auto_restart(true);
125 for (auto* h : worker_hosts) {
126 sg4::Actor::create("worker", h, worker, id, mailbox, std::ref(todo))->set_auto_restart(true);
132 XBT_INFO("WE SURVIVED!");