1 // -*- mode:C++; tab-width:8; c-basic-offset:2; indent-tabs-mode:t -*-
2 // vim: ts=8 sw=2 smarttab
4 * Ceph - scalable distributed file system
6 * Copyright (C) 2014 Adam Crume <adamcrume@gmail.com>
8 * This is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License version 2.1, as published by the Free Software
11 * Foundation. See file COPYING.
15 #include "Replayer.hpp"
16 #include "common/errno.h"
17 #include "rbd_replay/ActionTypes.h"
18 #include "rbd_replay/BufferReader.h"
19 #include <boost/foreach.hpp>
20 #include <boost/thread/thread.hpp>
21 #include <boost/scope_exit.hpp>
23 #include "global/global_context.h"
24 #include "rbd_replay_debug.hpp"
26 #define dout_context g_ceph_context
29 using namespace rbd_replay;
33 bool is_versioned_replay(BufferReader &buffer_reader) {
34 bufferlist::iterator *it;
35 int r = buffer_reader.fetch(&it);
40 if (it->get_remaining() < action::BANNER.size()) {
45 it->copy(action::BANNER.size(), banner);
46 bool versioned = (banner == action::BANNER);
53 } // anonymous namespace
55 Worker::Worker(Replayer &replayer)
56 : m_replayer(replayer),
61 void Worker::start() {
62 m_thread = boost::shared_ptr<boost::thread>(new boost::thread(boost::bind(&Worker::run, this)));
65 // Should only be called by StopThreadAction
74 void Worker::send(Action::ptr action) {
76 m_buffer.push_front(action);
79 void Worker::add_pending(PendingIO::ptr io) {
81 boost::mutex::scoped_lock lock(m_pending_ios_mutex);
82 assertf(m_pending_ios.count(io->id()) == 0, "id = %d", io->id());
83 m_pending_ios[io->id()] = io;
87 dout(THREAD_LEVEL) << "Worker thread started" << dendl;
90 m_buffer.pop_back(&action);
91 m_replayer.wait_for_actions(action->predecessors());
92 action->perform(*this);
93 m_replayer.set_action_complete(action->id());
96 boost::mutex::scoped_lock lock(m_pending_ios_mutex);
97 bool first_time = true;
98 while (!m_pending_ios.empty()) {
100 dout(THREAD_LEVEL) << "Worker thread trying to stop, still waiting for " << m_pending_ios.size() << " pending IOs to complete:" << dendl;
101 pair<action_id_t, PendingIO::ptr> p;
102 BOOST_FOREACH(p, m_pending_ios) {
103 dout(THREAD_LEVEL) << "> " << p.first << dendl;
106 m_pending_ios_empty.timed_wait(lock, boost::posix_time::seconds(1));
110 dout(THREAD_LEVEL) << "Worker thread stopped" << dendl;
114 void Worker::remove_pending(PendingIO::ptr io) {
116 m_replayer.set_action_complete(io->id());
117 boost::mutex::scoped_lock lock(m_pending_ios_mutex);
118 size_t num_erased = m_pending_ios.erase(io->id());
119 assertf(num_erased == 1, "id = %d", io->id());
120 if (m_pending_ios.empty()) {
121 m_pending_ios_empty.notify_all();
126 librbd::Image* Worker::get_image(imagectx_id_t imagectx_id) {
127 return m_replayer.get_image(imagectx_id);
131 void Worker::put_image(imagectx_id_t imagectx_id, librbd::Image* image) {
133 m_replayer.put_image(imagectx_id, image);
137 void Worker::erase_image(imagectx_id_t imagectx_id) {
138 m_replayer.erase_image(imagectx_id);
142 librbd::RBD* Worker::rbd() {
143 return m_replayer.get_rbd();
147 librados::IoCtx* Worker::ioctx() {
148 return m_replayer.get_ioctx();
151 void Worker::set_action_complete(action_id_t id) {
152 m_replayer.set_action_complete(id);
155 bool Worker::readonly() const {
156 return m_replayer.readonly();
159 rbd_loc Worker::map_image_name(string image_name, string snap_name) const {
160 return m_replayer.image_name_map().map(rbd_loc("", image_name, snap_name));
164 Replayer::Replayer(int num_action_trackers)
165 : m_rbd(NULL), m_ioctx(0),
166 m_latency_multiplier(1.0),
167 m_readonly(false), m_dump_perf_counters(false),
168 m_num_action_trackers(num_action_trackers),
169 m_action_trackers(new action_tracker_d[m_num_action_trackers]) {
170 assertf(num_action_trackers > 0, "num_action_trackers = %d", num_action_trackers);
173 Replayer::~Replayer() {
174 delete[] m_action_trackers;
177 Replayer::action_tracker_d &Replayer::tracker_for(action_id_t id) {
178 return m_action_trackers[id % m_num_action_trackers];
181 void Replayer::run(const std::string& replay_file) {
183 librados::Rados rados;
185 int r = rados.init_with_context(g_ceph_context);
187 cerr << "Failed to initialize RADOS: " << cpp_strerror(r) << std::endl;
192 cerr << "Failed to connect to cluster: " << cpp_strerror(r) << std::endl;
196 if (m_pool_name.empty()) {
197 r = rados.conf_get("rbd_default_pool", m_pool_name);
199 cerr << "Failed to retrieve default pool: " << cpp_strerror(r)
205 m_ioctx = new librados::IoCtx();
207 r = rados.ioctx_create(m_pool_name.c_str(), *m_ioctx);
209 cerr << "Failed to open pool " << m_pool_name << ": "
210 << cpp_strerror(r) << std::endl;
213 m_rbd = new librbd::RBD();
214 map<thread_id_t, Worker*> workers;
216 int fd = open(replay_file.c_str(), O_RDONLY);
218 std::cerr << "Failed to open " << replay_file << ": "
219 << cpp_strerror(errno) << std::endl;
222 BOOST_SCOPE_EXIT( (fd) ) {
224 } BOOST_SCOPE_EXIT_END;
226 BufferReader buffer_reader(fd);
227 bool versioned = is_versioned_replay(buffer_reader);
229 action::ActionEntry action_entry;
231 bufferlist::iterator *it;
232 int r = buffer_reader.fetch(&it);
234 std::cerr << "Failed to read from trace file: " << cpp_strerror(r)
238 if (it->get_remaining() == 0) {
243 action_entry.decode(*it);
245 action_entry.decode_unversioned(*it);
247 } catch (const buffer::error &err) {
248 std::cerr << "Failed to decode trace action: " << err.what() << std::endl;
252 Action::ptr action = Action::construct(action_entry);
254 // unknown / unsupported action
258 if (action->is_start_thread()) {
259 Worker *worker = new Worker(*this);
260 workers[action->thread_id()] = worker;
263 workers[action->thread_id()]->send(action);
267 dout(THREAD_LEVEL) << "Waiting for workers to die" << dendl;
268 pair<thread_id_t, Worker*> w;
269 BOOST_FOREACH(w, workers) {
286 librbd::Image* Replayer::get_image(imagectx_id_t imagectx_id) {
287 boost::shared_lock<boost::shared_mutex> lock(m_images_mutex);
288 return m_images[imagectx_id];
291 void Replayer::put_image(imagectx_id_t imagectx_id, librbd::Image *image) {
293 boost::unique_lock<boost::shared_mutex> lock(m_images_mutex);
294 assert(m_images.count(imagectx_id) == 0);
295 m_images[imagectx_id] = image;
298 void Replayer::erase_image(imagectx_id_t imagectx_id) {
299 boost::unique_lock<boost::shared_mutex> lock(m_images_mutex);
300 librbd::Image* image = m_images[imagectx_id];
301 if (m_dump_perf_counters) {
302 string command = "perf dump";
304 string format = "json-pretty";
306 g_ceph_context->do_command(command, cmdmap, format, &out);
307 out.write_stream(cout);
312 m_images.erase(imagectx_id);
315 void Replayer::set_action_complete(action_id_t id) {
316 dout(DEPGRAPH_LEVEL) << "ActionTracker::set_complete(" << id << ")" << dendl;
317 boost::system_time now(boost::get_system_time());
318 action_tracker_d &tracker = tracker_for(id);
319 boost::unique_lock<boost::shared_mutex> lock(tracker.mutex);
320 assert(tracker.actions.count(id) == 0);
321 tracker.actions[id] = now;
322 tracker.condition.notify_all();
325 bool Replayer::is_action_complete(action_id_t id) {
326 action_tracker_d &tracker = tracker_for(id);
327 boost::shared_lock<boost::shared_mutex> lock(tracker.mutex);
328 return tracker.actions.count(id) > 0;
331 void Replayer::wait_for_actions(const action::Dependencies &deps) {
332 boost::posix_time::ptime release_time(boost::posix_time::neg_infin);
333 BOOST_FOREACH(const action::Dependency &dep, deps) {
334 dout(DEPGRAPH_LEVEL) << "Waiting for " << dep.id << dendl;
335 boost::system_time start_time(boost::get_system_time());
336 action_tracker_d &tracker = tracker_for(dep.id);
337 boost::shared_lock<boost::shared_mutex> lock(tracker.mutex);
338 bool first_time = true;
339 while (tracker.actions.count(dep.id) == 0) {
341 dout(DEPGRAPH_LEVEL) << "Still waiting for " << dep.id << dendl;
343 tracker.condition.timed_wait(lock, boost::posix_time::seconds(1));
346 boost::system_time action_completed_time(tracker.actions[dep.id]);
348 boost::system_time end_time(boost::get_system_time());
349 long long micros = (end_time - start_time).total_microseconds();
350 dout(DEPGRAPH_LEVEL) << "Finished waiting for " << dep.id << " after " << micros << " microseconds" << dendl;
351 // Apparently the nanoseconds constructor is optional:
352 // http://www.boost.org/doc/libs/1_46_0/doc/html/date_time/details.html#compile_options
353 boost::system_time sub_release_time(action_completed_time + boost::posix_time::microseconds(dep.time_delta * m_latency_multiplier / 1000));
354 if (sub_release_time > release_time) {
355 release_time = sub_release_time;
358 if (release_time > boost::get_system_time()) {
359 dout(SLEEP_LEVEL) << "Sleeping for " << (release_time - boost::get_system_time()).total_microseconds() << " microseconds" << dendl;
360 boost::this_thread::sleep(release_time);
364 void Replayer::clear_images() {
365 boost::unique_lock<boost::shared_mutex> lock(m_images_mutex);
366 if (m_dump_perf_counters && !m_images.empty()) {
367 string command = "perf dump";
369 string format = "json-pretty";
371 g_ceph_context->do_command(command, cmdmap, format, &out);
372 out.write_stream(cout);
376 pair<imagectx_id_t, librbd::Image*> p;
377 BOOST_FOREACH(p, m_images) {
383 void Replayer::set_latency_multiplier(float f) {
384 assertf(f >= 0, "f = %f", f);
385 m_latency_multiplier = f;
388 bool Replayer::readonly() const {
392 void Replayer::set_readonly(bool readonly) {
393 m_readonly = readonly;
396 string Replayer::pool_name() const {
400 void Replayer::set_pool_name(string pool_name) {
401 m_pool_name = pool_name;