// Handle the reply from a scheduler // int CLIENT_STATE::handle_scheduler_reply( PROJECT* project, char* scheduler_url ) { SCHEDULER_REPLY sr; FILE* f; int retval; unsigned int i; bool signature_valid, update_global_prefs=false, update_project_prefs=false; char buf[1024], filename[256]; std::string old_gui_urls = project->gui_urls; PROJECT* p2; vector<RESULT*>new_results; project->last_rpc_time = now; if (requested_work()) { had_or_requested_work = true; } get_sched_reply_filename(*project, filename, sizeof(filename)); f = fopen(filename, "r"); if (!f) return ERR_FOPEN; retval = sr.parse(f, project); fclose(f); if (retval) return retval; if (log_flags.sched_ops) { if (requested_work()) { sprintf(buf, ": got %d new tasks", (int)sr.results.size()); } else { strcpy(buf, ""); } msg_printf(project, MSG_INFO, "Scheduler request completed%s", buf); } if (log_flags.sched_op_debug) { if (sr.scheduler_version) { msg_printf(project, MSG_INFO, "[sched_op] Server version %d", sr.scheduler_version ); } } // check that master URL is correct // if (strlen(sr.master_url)) { canonicalize_master_url(sr.master_url); string url1 = sr.master_url; string url2 = project->master_url; downcase_string(url1); downcase_string(url2); if (url1 != url2) { p2 = lookup_project(sr.master_url); if (p2) { msg_printf(project, MSG_USER_ALERT, "You are attached to this project twice. Please remove projects named %s, then add %s", project->project_name, sr.master_url ); } else { msg_printf(project, MSG_INFO, _("You used the wrong URL for this project. When convenient, remove this project, then add %s"), sr.master_url ); } } } // make sure we don't already have a project of same name // bool dup_name = false; for (i=0; i<projects.size(); i++) { p2 = projects[i]; if (project == p2) continue; if (!strcmp(p2->project_name, project->project_name)) { dup_name = true; break; } } if (dup_name) { msg_printf(project, MSG_INFO, "Already attached to a project named %s (possibly with wrong URL)", project->project_name ); msg_printf(project, MSG_INFO, "Consider detaching this project, then trying again" ); } // show messages from server // for (i=0; i<sr.messages.size(); i++) { USER_MESSAGE& um = sr.messages[i]; int prio = (!strcmp(um.priority.c_str(), "notice"))?MSG_SCHEDULER_ALERT:MSG_INFO; string_substitute(um.message.c_str(), buf, sizeof(buf), "%", "%%"); msg_printf(project, prio, "%s", buf); } if (log_flags.sched_op_debug && sr.request_delay) { msg_printf(project, MSG_INFO, "Project requested delay of %.0f seconds", sr.request_delay ); } // if project is down, return error (so that we back off) // and don't do anything else // if (sr.project_is_down) { if (sr.request_delay) { double x = now + sr.request_delay; project->set_min_rpc_time(x, "project is down"); } return ERR_PROJECT_DOWN; } // if the scheduler reply includes global preferences, // insert extra elements, write to disk, and parse // if (sr.global_prefs_xml) { // skip this if we have host-specific prefs // and we're talking to an old scheduler // if (!global_prefs.host_specific || sr.scheduler_version >= 507) { retval = save_global_prefs( sr.global_prefs_xml, project->master_url, scheduler_url ); if (retval) { return retval; } update_global_prefs = true; } else { if (log_flags.sched_op_debug) { msg_printf(project, MSG_INFO, "ignoring prefs from old server; we have host-specific prefs" ); } } } // see if we have a new venue from this project // (this must go AFTER the above, since otherwise // global_prefs_source_project() is meaningless) // if (strcmp(project->host_venue, sr.host_venue)) { safe_strcpy(project->host_venue, sr.host_venue); msg_printf(project, MSG_INFO, "New computer location: %s", sr.host_venue); update_project_prefs = true; if (project == global_prefs_source_project()) { strcpy(main_host_venue, sr.host_venue); update_global_prefs = true; } } if (update_global_prefs) { read_global_prefs(); } // deal with project preferences (should always be there) // If they've changed, write to account file, // then parse to get our venue, and pass to running apps // if (sr.project_prefs_xml) { if (strcmp(project->project_prefs.c_str(), sr.project_prefs_xml)) { project->project_prefs = string(sr.project_prefs_xml); update_project_prefs = true; } } // the account file has GUI URLs and project prefs. // rewrite if either of these has changed // if (project->gui_urls != old_gui_urls || update_project_prefs) { retval = project->write_account_file(); if (retval) { msg_printf(project, MSG_INTERNAL_ERROR, "Can't write account file: %s", boincerror(retval) ); return retval; } } if (update_project_prefs) { project->parse_account_file(); if (strlen(project->host_venue)) { project->parse_account_file_venue(); } project->parse_preferences_for_user_files(); active_tasks.request_reread_prefs(project); } // if the scheduler reply includes a code-signing key, // accept it if we don't already have one from the project. // Otherwise verify its signature, using the key we already have. // if (sr.code_sign_key) { if (!strlen(project->code_sign_key)) { safe_strcpy(project->code_sign_key, sr.code_sign_key); } else { if (sr.code_sign_key_signature) { retval = check_string_signature2( sr.code_sign_key, sr.code_sign_key_signature, project->code_sign_key, signature_valid ); if (!retval && signature_valid) { safe_strcpy(project->code_sign_key, sr.code_sign_key); } else { msg_printf(project, MSG_INTERNAL_ERROR, "New code signing key doesn't validate" ); } } else { msg_printf(project, MSG_INTERNAL_ERROR, "Missing code sign key signature" ); } } } // copy new entities to client state // for (i=0; i<sr.apps.size(); i++) { APP* app = lookup_app(project, sr.apps[i].name); if (app) { strcpy(app->user_friendly_name, sr.apps[i].user_friendly_name); } else { app = new APP; *app = sr.apps[i]; retval = link_app(project, app); if (retval) { msg_printf(project, MSG_INTERNAL_ERROR, "Can't handle application %s in scheduler reply", app->name ); delete app; } else { apps.push_back(app); } } } FILE_INFO* fip; for (i=0; i<sr.file_infos.size(); i++) { fip = lookup_file_info(project, sr.file_infos[i].name); if (fip) { fip->merge_info(sr.file_infos[i]); } else { fip = new FILE_INFO; *fip = sr.file_infos[i]; retval = link_file_info(project, fip); if (retval) { msg_printf(project, MSG_INTERNAL_ERROR, "Can't handle file %s in scheduler reply", fip->name ); delete fip; } else { file_infos.push_back(fip); } } } for (i=0; i<sr.file_deletes.size(); i++) { fip = lookup_file_info(project, sr.file_deletes[i].c_str()); if (fip) { if (log_flags.file_xfer_debug) { msg_printf(project, MSG_INFO, "[file_xfer_debug] Got server request to delete file %s", fip->name ); } fip->sticky = false; } } for (i=0; i<sr.app_versions.size(); i++) { if (project->anonymous_platform) { msg_printf(project, MSG_INTERNAL_ERROR, "App version returned from anonymous platform project; ignoring" ); continue; } APP_VERSION& avpp = sr.app_versions[i]; if (strlen(avpp.platform) == 0) { strcpy(avpp.platform, get_primary_platform()); } else { if (!is_supported_platform(avpp.platform)) { msg_printf(project, MSG_INTERNAL_ERROR, "App version has unsupported platform %s", avpp.platform ); continue; } } if (avpp.missing_coproc) { msg_printf(project, MSG_INTERNAL_ERROR, "App version uses non-existent %s GPU", avpp.missing_coproc_name ); } APP* app = lookup_app(project, avpp.app_name); if (!app) { msg_printf(project, MSG_INTERNAL_ERROR, "Missing app %s", avpp.app_name ); continue; } APP_VERSION* avp = lookup_app_version( app, avpp.platform, avpp.version_num, avpp.plan_class ); if (avp) { // update performance-related info; // generally this shouldn't change, // but if it does it's better to use the new stuff // avp->avg_ncpus = avpp.avg_ncpus; avp->max_ncpus = avpp.max_ncpus; avp->flops = avpp.flops; strcpy(avp->cmdline, avpp.cmdline); avp->gpu_usage = avpp.gpu_usage; strlcpy(avp->api_version, avpp.api_version, sizeof(avp->api_version)); avp->dont_throttle = avpp.dont_throttle; avp->needs_network = avpp.needs_network; // if we had download failures, clear them // avp->clear_errors(); continue; } avp = new APP_VERSION; *avp = avpp; retval = link_app_version(project, avp); if (retval) { delete avp; continue; } app_versions.push_back(avp); } for (i=0; i<sr.workunits.size(); i++) { if (lookup_workunit(project, sr.workunits[i].name)) continue; WORKUNIT* wup = new WORKUNIT; *wup = sr.workunits[i]; wup->project = project; retval = link_workunit(project, wup); if (retval) { msg_printf(project, MSG_INTERNAL_ERROR, "Can't handle task %s in scheduler reply", wup->name ); delete wup; continue; } wup->clear_errors(); workunits.push_back(wup); } double est_rsc_runtime[MAX_RSC]; for (int j=0; j<coprocs.n_rsc; j++) { est_rsc_runtime[j] = 0; } for (i=0; i<sr.results.size(); i++) { if (lookup_result(project, sr.results[i].name)) { msg_printf(project, MSG_INTERNAL_ERROR, "Already have task %s\n", sr.results[i].name ); continue; } RESULT* rp = new RESULT; *rp = sr.results[i]; retval = link_result(project, rp); if (retval) { msg_printf(project, MSG_INTERNAL_ERROR, "Can't handle task %s in scheduler reply", rp->name ); delete rp; continue; } if (strlen(rp->platform) == 0) { strcpy(rp->platform, get_primary_platform()); rp->version_num = latest_version(rp->wup->app, rp->platform); } rp->avp = lookup_app_version( rp->wup->app, rp->platform, rp->version_num, rp->plan_class ); if (!rp->avp) { msg_printf(project, MSG_INTERNAL_ERROR, "No app version found for app %s platform %s ver %d class %s; discarding %s", rp->wup->app->name, rp->platform, rp->version_num, rp->plan_class, rp->name ); delete rp; continue; } if (rp->avp->missing_coproc) { msg_printf(project, MSG_INTERNAL_ERROR, "Missing coprocessor for task %s; aborting", rp->name ); rp->abort_inactive(EXIT_MISSING_COPROC); } else { rp->set_state(RESULT_NEW, "handle_scheduler_reply"); int rt = rp->avp->gpu_usage.rsc_type; if (rt > 0) { est_rsc_runtime[rt] += rp->estimated_runtime(); gpus_usable = true; // trigger a check of whether GPU is actually usable } else { est_rsc_runtime[0] += rp->estimated_runtime(); } } rp->wup->version_num = rp->version_num; rp->received_time = now; new_results.push_back(rp); results.push_back(rp); } sort_results(); if (log_flags.sched_op_debug) { if (sr.results.size()) { for (int j=0; j<coprocs.n_rsc; j++) { msg_printf(project, MSG_INFO, "[sched_op] estimated total %s task duration: %.0f seconds", rsc_name(j), est_rsc_runtime[j]/time_stats.availability_frac(j) ); } } } // update records for ack'ed results // for (i=0; i<sr.result_acks.size(); i++) { if (log_flags.sched_op_debug) { msg_printf(project, MSG_INFO, "[sched_op] handle_scheduler_reply(): got ack for task %s\n", sr.result_acks[i].name ); } RESULT* rp = lookup_result(project, sr.result_acks[i].name); if (rp) { rp->got_server_ack = true; } else { msg_printf(project, MSG_INTERNAL_ERROR, "Got ack for task %s, but can't find it", sr.result_acks[i].name ); } } // handle result abort requests // for (i=0; i<sr.result_abort.size(); i++) { RESULT* rp = lookup_result(project, sr.result_abort[i].name); if (rp) { ACTIVE_TASK* atp = lookup_active_task_by_result(rp); if (atp) { atp->abort_task(EXIT_ABORTED_BY_PROJECT, "aborted by project - no longer usable" ); } else { rp->abort_inactive(EXIT_ABORTED_BY_PROJECT); } } else { msg_printf(project, MSG_INTERNAL_ERROR, "Server requested abort of unknown task %s", sr.result_abort[i].name ); } } for (i=0; i<sr.result_abort_if_not_started.size(); i++) { RESULT* rp = lookup_result(project, sr.result_abort_if_not_started[i].name); if (!rp) { msg_printf(project, MSG_INTERNAL_ERROR, "Server requested conditional abort of unknown task %s", sr.result_abort_if_not_started[i].name ); continue; } if (rp->not_started) { rp->abort_inactive(EXIT_ABORTED_BY_PROJECT); } } // remove acked trickle files // if (sr.message_ack) { remove_trickle_files(project); } if (sr.send_full_workload) { project->send_full_workload = true; } project->dont_use_dcf = sr.dont_use_dcf; project->send_time_stats_log = sr.send_time_stats_log; project->send_job_log = sr.send_job_log; project->trickle_up_pending = false; // The project returns a hostid only if it has created a new host record. // In that case reset RPC seqno // if (sr.hostid) { if (project->hostid) { // if we already have a host ID for this project, // we must have sent it a stale seqno, // which usually means our state file was copied from another host. // So generate a new host CPID. // generate_new_host_cpid(); msg_printf(project, MSG_INFO, "Generated new computer cross-project ID: %s", host_info.host_cpid ); } //msg_printf(project, MSG_INFO, "Changing host ID from %d to %d", project->hostid, sr.hostid); project->hostid = sr.hostid; project->rpc_seqno = 0; } #ifdef ENABLE_AUTO_UPDATE if (sr.auto_update.present) { if (!sr.auto_update.validate_and_link(project)) { auto_update = sr.auto_update; } } #endif project->project_files = sr.project_files; project->link_project_files(); project->create_project_file_symlinks(); if (log_flags.state_debug) { msg_printf(project, MSG_INFO, "[state] handle_scheduler_reply(): State after handle_scheduler_reply():" ); print_summary(); } // the following must precede the backoff and request_delay checks, // since it overrides them // if (sr.next_rpc_delay) { project->next_rpc_time = now + sr.next_rpc_delay; } else { project->next_rpc_time = 0; } work_fetch.handle_reply(project, &sr, new_results); project->nrpc_failures = 0; project->min_rpc_time = 0; if (sr.request_delay) { double x = now + sr.request_delay; project->set_min_rpc_time(x, "requested by project"); } if (sr.got_rss_feeds) { handle_sr_feeds(sr.sr_feeds, project); } update_trickle_up_urls(project, sr.trickle_up_urls); // garbage collect in case the project sent us some irrelevant FILE_INFOs; // avoid starting transfers for them // gstate.garbage_collect_always(); return 0; }
// Write a scheduler request to a disk file, // to be sent to a scheduling server // int CLIENT_STATE::make_scheduler_request(PROJECT* p) { char buf[1024]; MIOFILE mf; unsigned int i; RESULT* rp; get_sched_request_filename(*p, buf, sizeof(buf)); FILE* f = boinc_fopen(buf, "wb"); if (!f) return ERR_FOPEN; double trs = total_resource_share(); double rrs = runnable_resource_share(RSC_TYPE_ANY); double prrs = potentially_runnable_resource_share(); double resource_share_fraction, rrs_fraction, prrs_fraction; if (trs) { resource_share_fraction = p->resource_share / trs; } else { resource_share_fraction = 1; } if (rrs) { rrs_fraction = p->resource_share / rrs; } else { rrs_fraction = 1; } if (prrs) { prrs_fraction = p->resource_share / prrs; } else { prrs_fraction = 1; } // if hostid is zero, rpc_seqno better be also // if (!p->hostid) { p->rpc_seqno = 0; } mf.init_file(f); fprintf(f, "<scheduler_request>\n" " <authenticator>%s</authenticator>\n" " <hostid>%d</hostid>\n" " <rpc_seqno>%d</rpc_seqno>\n" " <core_client_major_version>%d</core_client_major_version>\n" " <core_client_minor_version>%d</core_client_minor_version>\n" " <core_client_release>%d</core_client_release>\n" " <resource_share_fraction>%f</resource_share_fraction>\n" " <rrs_fraction>%f</rrs_fraction>\n" " <prrs_fraction>%f</prrs_fraction>\n" " <duration_correction_factor>%f</duration_correction_factor>\n" " <allow_multiple_clients>%d</allow_multiple_clients>\n" " <sandbox>%d</sandbox>\n", p->authenticator, p->hostid, p->rpc_seqno, core_client_version.major, core_client_version.minor, core_client_version.release, resource_share_fraction, rrs_fraction, prrs_fraction, p->duration_correction_factor, config.allow_multiple_clients?1:0, g_use_sandbox?1:0 ); work_fetch.write_request(f, p); // write client capabilities // fprintf(f, " <client_cap_plan_class>1</client_cap_plan_class>\n" ); write_platforms(p, mf); if (strlen(p->code_sign_key)) { fprintf(f, " <code_sign_key>\n%s\n</code_sign_key>\n", p->code_sign_key); } // send working prefs // fprintf(f, "<working_global_preferences>\n"); global_prefs.write(mf); fprintf(f, "</working_global_preferences>\n"); // send master global preferences if present and not host-specific // if (!global_prefs.host_specific && boinc_file_exists(GLOBAL_PREFS_FILE_NAME)) { FILE* fprefs = fopen(GLOBAL_PREFS_FILE_NAME, "r"); if (fprefs) { copy_stream(fprefs, f); fclose(fprefs); } PROJECT* pp = lookup_project(global_prefs.source_project); if (pp && strlen(pp->email_hash)) { fprintf(f, "<global_prefs_source_email_hash>%s</global_prefs_source_email_hash>\n", pp->email_hash ); } } // Of the projects with same email hash as this one, // send the oldest cross-project ID. // Use project URL as tie-breaker. // PROJECT* winner = p; for (i=0; i<projects.size(); i++ ) { PROJECT* project = projects[i]; if (project == p) continue; if (strcmp(project->email_hash, p->email_hash)) continue; if (project->cpid_time < winner->cpid_time) { winner = project; } else if (project->cpid_time == winner->cpid_time) { if (strcmp(project->master_url, winner->master_url) < 0) { winner = project; } } } fprintf(f, "<cross_project_id>%s</cross_project_id>\n", winner->cross_project_id ); time_stats.write(mf, true); net_stats.write(mf); if (global_prefs.daily_xfer_period_days) { daily_xfer_history.write_scheduler_request( mf, global_prefs.daily_xfer_period_days ); } // update hardware info, and write host info // host_info.get_host_info(); set_ncpus(); host_info.write(mf, !config.suppress_net_info, false); // get and write disk usage // get_disk_usages(); get_disk_shares(); fprintf(f, " <disk_usage>\n" " <d_boinc_used_total>%f</d_boinc_used_total>\n" " <d_boinc_used_project>%f</d_boinc_used_project>\n" " <d_project_share>%f</d_project_share>\n" " </disk_usage>\n", total_disk_usage, p->disk_usage, p->disk_share ); // copy request values from RSC_WORK_FETCH to COPROC // int j = rsc_index(GPU_TYPE_NVIDIA); if (j > 0) { coprocs.nvidia.req_secs = rsc_work_fetch[j].req_secs; coprocs.nvidia.req_instances = rsc_work_fetch[j].req_instances; coprocs.nvidia.estimated_delay = rsc_work_fetch[j].req_secs?rsc_work_fetch[j].busy_time_estimator.get_busy_time():0; } j = rsc_index(GPU_TYPE_ATI); if (j > 0) { coprocs.ati.req_secs = rsc_work_fetch[j].req_secs; coprocs.ati.req_instances = rsc_work_fetch[j].req_instances; coprocs.ati.estimated_delay = rsc_work_fetch[j].req_secs?rsc_work_fetch[j].busy_time_estimator.get_busy_time():0; } if (coprocs.n_rsc > 1) { coprocs.write_xml(mf, true); } // report completed jobs // unsigned int last_reported_index = 0; p->nresults_returned = 0; for (i=0; i<results.size(); i++) { rp = results[i]; if (rp->project == p && rp->ready_to_report) { p->nresults_returned++; rp->write(mf, true); } if (config.max_tasks_reported && (p->nresults_returned >= config.max_tasks_reported) ) { last_reported_index = i; break; } } read_trickle_files(p, f); // report sticky files as needed // for (i=0; i<file_infos.size(); i++) { FILE_INFO* fip = file_infos[i]; if (fip->project != p) continue; if (!fip->sticky) continue; fprintf(f, " <file_info>\n" " <name>%s</name>\n" " <nbytes>%f</nbytes>\n" " <status>%d</status>\n" " </file_info>\n", fip->name, fip->nbytes, fip->status ); } if (p->send_time_stats_log) { fprintf(f, "<time_stats_log>\n"); time_stats.get_log_after(p->send_time_stats_log, mf); fprintf(f, "</time_stats_log>\n"); } if (p->send_job_log) { fprintf(f, "<job_log>\n"); job_log_filename(*p, buf, sizeof(buf)); send_log_after(buf, p->send_job_log, mf); fprintf(f, "</job_log>\n"); } // send descriptions of app versions // fprintf(f, "<app_versions>\n"); j=0; for (i=0; i<app_versions.size(); i++) { APP_VERSION* avp = app_versions[i]; if (avp->project != p) continue; avp->write(mf, false); avp->index = j++; } fprintf(f, "</app_versions>\n"); // send descriptions of jobs in progress for this project // fprintf(f, "<other_results>\n"); for (i=0; i<results.size(); i++) { rp = results[i]; if (rp->project != p) continue; if ((last_reported_index && (i > last_reported_index)) || !rp->ready_to_report) { fprintf(f, " <other_result>\n" " <name>%s</name>\n" " <app_version>%d</app_version>\n", rp->name, rp->avp->index ); // the following is for backwards compatibility w/ old schedulers // if (strlen(rp->avp->plan_class)) { fprintf(f, " <plan_class>%s</plan_class>\n", rp->avp->plan_class ); } fprintf(f, " </other_result>\n" ); } } fprintf(f, "</other_results>\n"); // if requested by project, send summary of all in-progress results // (for EDF simulation by scheduler) // if (p->send_full_workload) { fprintf(f, "<in_progress_results>\n"); for (i=0; i<results.size(); i++) { rp = results[i]; double x = rp->estimated_runtime_remaining(); if (x == 0) continue; strcpy(buf, ""); int rt = rp->avp->gpu_usage.rsc_type; if (rt) { if (rt == rsc_index(GPU_TYPE_NVIDIA)) { sprintf(buf, " <ncudas>%f</ncudas>\n", rp->avp->gpu_usage.usage); } else if (rt == rsc_index(GPU_TYPE_ATI)) { sprintf(buf, " <natis>%f</natis>\n", rp->avp->gpu_usage.usage); } } fprintf(f, " <ip_result>\n" " <name>%s</name>\n" " <report_deadline>%.0f</report_deadline>\n" " <time_remaining>%.2f</time_remaining>\n" " <avg_ncpus>%f</avg_ncpus>\n" "%s" " </ip_result>\n", rp->name, rp->report_deadline, x, rp->avp->avg_ncpus, buf ); } fprintf(f, "</in_progress_results>\n"); } FILE* cof = boinc_fopen(CLIENT_OPAQUE_FILENAME, "r"); if (cof) { fprintf(f, "<client_opaque>\n<![CDATA[\n"); copy_stream(cof, f); fprintf(f, "\n]]>\n</client_opaque>\n"); fclose(cof); } fprintf(f, "</scheduler_request>\n"); fclose(f); return 0; }
// parse a scheduler reply. // Some of the items go into the SCHEDULER_REPLY object. // Others are copied straight to the PROJECT // int SCHEDULER_REPLY::parse(FILE* in, PROJECT* project) { char buf[256], msg_buf[1024], pri_buf[256], attr_buf[256]; int retval; MIOFILE mf; XML_PARSER xp(&mf); std::string delete_file_name; mf.init_file(in); bool found_start_tag = false, btemp; double cpid_time = 0; clear(); safe_strcpy(host_venue, project->host_venue); // the project won't send us a venue if it's doing maintenance // or doesn't check the DB because no work. // Don't overwrite the host venue in that case. sr_feeds.clear(); trickle_up_urls.clear(); if (!project->anonymous_platform) { for (int i=0; i<MAX_RSC; i++) { project->no_rsc_apps[i] = false; } } // First line should either be tag (HTTP 1.0) or // hex length of response (HTTP 1.1) // while (!xp.get_tag(attr_buf, sizeof(attr_buf))) { if (!found_start_tag) { if (xp.match_tag("scheduler_reply")) { found_start_tag = true; } continue; } if (xp.match_tag("/scheduler_reply")) { // update statistics after parsing the scheduler reply // add new record if vector is empty or we have a new day // if (project->statistics.empty() || project->statistics.back().day!=dday()) { project->trim_statistics(); DAILY_STATS nds; project->statistics.push_back(nds); } DAILY_STATS& ds = project->statistics.back(); ds.day=dday(); ds.user_total_credit=project->user_total_credit; ds.user_expavg_credit=project->user_expavg_credit; ds.host_total_credit=project->host_total_credit; ds.host_expavg_credit=project->host_expavg_credit; project->write_statistics_file(); if (cpid_time) { project->cpid_time = cpid_time; } else { project->cpid_time = project->user_create_time; } if (project->dont_use_dcf) { project->duration_correction_factor = 1; } return 0; } else if (xp.parse_str("project_name", project->project_name, sizeof(project->project_name))) { continue; } else if (xp.parse_str("master_url", master_url, sizeof(master_url))) { continue; } else if (xp.parse_str("symstore", project->symstore, sizeof(project->symstore))) continue; else if (xp.parse_str("user_name", project->user_name, sizeof(project->user_name))) continue; else if (xp.parse_double("user_total_credit", project->user_total_credit)) continue; else if (xp.parse_double("user_expavg_credit", project->user_expavg_credit)) continue; else if (xp.parse_double("user_create_time", project->user_create_time)) continue; else if (xp.parse_double("cpid_time", cpid_time)) continue; else if (xp.parse_str("team_name", project->team_name, sizeof(project->team_name))) continue; else if (xp.parse_int("hostid", hostid)) continue; else if (xp.parse_double("host_total_credit", project->host_total_credit)) continue; else if (xp.parse_double("host_expavg_credit", project->host_expavg_credit)) continue; else if (xp.parse_str("host_venue", host_venue, sizeof(host_venue))) continue; else if (xp.parse_double("host_create_time", project->host_create_time)) continue; else if (xp.parse_double("request_delay", request_delay)) continue; else if (xp.parse_double("next_rpc_delay", next_rpc_delay)) continue; else if (xp.match_tag("global_preferences")) { retval = dup_element_contents( xp.f->f, "</global_preferences>", &global_prefs_xml ); if (retval) { msg_printf(project, MSG_INTERNAL_ERROR, "Can't parse global prefs in scheduler reply: %s", boincerror(retval) ); return retval; } } else if (xp.match_tag("project_preferences")) { retval = dup_element_contents( xp.f->f, "</project_preferences>", &project_prefs_xml ); if (retval) { msg_printf(project, MSG_INTERNAL_ERROR, "Can't parse project prefs in scheduler reply: %s", boincerror(retval) ); return retval; } } else if (xp.match_tag("gui_urls")) { std::string foo; retval = copy_element_contents(xp.f->f, "</gui_urls>", foo); if (retval) { msg_printf(project, MSG_INTERNAL_ERROR, "Can't parse GUI URLs in scheduler reply: %s", boincerror(retval) ); return retval; } project->gui_urls = "<gui_urls>\n"+foo+"</gui_urls>\n"; } else if (xp.match_tag("code_sign_key")) { retval = dup_element_contents( xp.f->f, "</code_sign_key>", &code_sign_key ); if (retval) { msg_printf(project, MSG_INTERNAL_ERROR, "Can't parse code sign key in scheduler reply: %s", boincerror(retval) ); return ERR_XML_PARSE; } strip_whitespace(code_sign_key); } else if (xp.match_tag("code_sign_key_signature")) { retval = dup_element_contents( xp.f->f, "</code_sign_key_signature>", &code_sign_key_signature ); if (retval) { msg_printf(project, MSG_INTERNAL_ERROR, "Can't parse code sign key signature in scheduler reply: %s", boincerror(retval) ); return ERR_XML_PARSE; } } else if (xp.match_tag("app")) { APP app; retval = app.parse(xp); if (retval) { msg_printf(project, MSG_INTERNAL_ERROR, "Can't parse application in scheduler reply: %s", boincerror(retval) ); } else { apps.push_back(app); } } else if (xp.match_tag("file_info")) { FILE_INFO file_info; retval = file_info.parse(xp); if (retval) { msg_printf(project, MSG_INTERNAL_ERROR, "Can't parse file info in scheduler reply: %s", boincerror(retval) ); } else { file_infos.push_back(file_info); } } else if (xp.match_tag("app_version")) { APP_VERSION av; retval = av.parse(xp); if (retval) { msg_printf(project, MSG_INTERNAL_ERROR, "Can't parse application version in scheduler reply: %s", boincerror(retval) ); } else { app_versions.push_back(av); } } else if (xp.match_tag("workunit")) { WORKUNIT wu; retval = wu.parse(xp); if (retval) { msg_printf(project, MSG_INTERNAL_ERROR, "Can't parse workunit in scheduler reply: %s", boincerror(retval) ); } else { workunits.push_back(wu); } } else if (xp.match_tag("result")) { RESULT result; // make sure this is here so constructor // gets called each time retval = result.parse_server(xp); if (retval) { msg_printf(project, MSG_INTERNAL_ERROR, "Can't parse task in scheduler reply: %s", boincerror(retval) ); } else { results.push_back(result); } } else if (xp.match_tag("result_ack")) { RESULT result; retval = result.parse_name(xp, "/result_ack"); if (retval) { msg_printf(project, MSG_INTERNAL_ERROR, "Can't parse ack in scheduler reply: %s", boincerror(retval) ); } else { result_acks.push_back(result); } } else if (xp.match_tag("result_abort")) { RESULT result; retval = result.parse_name(xp, "/result_abort"); if (retval) { msg_printf(project, MSG_INTERNAL_ERROR, "Can't parse result abort in scheduler reply: %s", boincerror(retval) ); } else { result_abort.push_back(result); } } else if (xp.match_tag("result_abort_if_not_started")) { RESULT result; retval = result.parse_name(xp, "/result_abort_if_not_started"); if (retval) { msg_printf(project, MSG_INTERNAL_ERROR, "Can't parse result abort-if-not-started in scheduler reply: %s", boincerror(retval) ); } else { result_abort_if_not_started.push_back(result); } } else if (xp.parse_string("delete_file_info", delete_file_name)) { file_deletes.push_back(delete_file_name); } else if (xp.parse_str("message", msg_buf, sizeof(msg_buf))) { parse_attr(attr_buf, "priority", pri_buf, sizeof(pri_buf)); USER_MESSAGE um(msg_buf, pri_buf); messages.push_back(um); continue; } else if (xp.parse_bool("message_ack", message_ack)) { continue; } else if (xp.parse_bool("project_is_down", project_is_down)) { continue; } else if (xp.parse_str("email_hash", project->email_hash, sizeof(project->email_hash))) { continue; } else if (xp.parse_str("cross_project_id", project->cross_project_id, sizeof(project->cross_project_id))) { continue; } else if (xp.parse_str("external_cpid", project->external_cpid, sizeof(project->external_cpid))) { continue; } else if (xp.match_tag("trickle_down")) { retval = gstate.handle_trickle_down(project, in); if (retval) { msg_printf(project, MSG_INTERNAL_ERROR, "handle_trickle_down failed: %s", boincerror(retval) ); } continue; } else if (xp.parse_bool("non_cpu_intensive", project->non_cpu_intensive)) { continue; } else if (xp.parse_bool("ended", project->ended)) { continue; } else if (xp.parse_bool("no_cpu_apps", btemp)) { if (!project->anonymous_platform) { handle_no_rsc_apps("CPU", project, btemp); } continue; // deprecated syntax } else if (xp.parse_bool("no_cuda_apps", btemp)) { if (!project->anonymous_platform) { handle_no_rsc_apps(GPU_TYPE_NVIDIA, project, btemp); } continue; } else if (xp.parse_bool("no_ati_apps", btemp)) { if (!project->anonymous_platform) { handle_no_rsc_apps(GPU_TYPE_ATI, project, btemp); } continue; } else if (xp.parse_str("no_rsc_apps", buf, sizeof(buf))) { if (!project->anonymous_platform) { handle_no_rsc_apps(buf, project, true); } continue; } else if (xp.parse_bool("verify_files_on_app_start", project->verify_files_on_app_start)) { continue; } else if (xp.parse_bool("send_full_workload", send_full_workload)) { continue; } else if (xp.parse_bool("dont_use_dcf", dont_use_dcf)) { continue; } else if (xp.parse_int("send_time_stats_log", send_time_stats_log)){ continue; } else if (xp.parse_int("send_job_log", send_job_log)) { continue; } else if (xp.parse_int("scheduler_version", scheduler_version)) { continue; } else if (xp.match_tag("project_files")) { retval = parse_project_files(xp, project_files); #ifdef ENABLE_AUTO_UPDATE } else if (xp.match_tag("auto_update")) { retval = auto_update.parse(xp); if (!retval) auto_update.present = true; #endif } else if (xp.match_tag("rss_feeds")) { got_rss_feeds = true; parse_rss_feed_descs(xp, sr_feeds); continue; } else if (xp.match_tag("trickle_up_urls")) { parse_trickle_up_urls(xp, trickle_up_urls); continue; } else if (xp.parse_int("userid", project->userid)) { continue; } else if (xp.parse_int("teamid", project->teamid)) { continue; } else if (xp.parse_double("desired_disk_usage", project->desired_disk_usage)) { continue; } else { if (log_flags.unparsed_xml) { msg_printf(project, MSG_INFO, "[unparsed_xml] SCHEDULER_REPLY::parse(): unrecognized %s\n", xp.parsed_tag ); } } } if (found_start_tag) { msg_printf(project, MSG_INTERNAL_ERROR, "No close tag in scheduler reply"); } else { msg_printf(project, MSG_INTERNAL_ERROR, "No start tag in scheduler reply"); } return ERR_XML_PARSE; }
// The following runs "test_app" and sends it various messages. // Used for testing the runtime system. // void run_test_app() { WORKUNIT wu; PROJECT project; APP app; APP_VERSION av; ACTIVE_TASK at; ACTIVE_TASK_SET ats; RESULT result; int retval; char buf[256]; getcwd(buf, sizeof(buf)); // so we can see where we're running gstate.run_test_app = true; wu.project = &project; wu.app = &app; wu.command_line = string("--critical_section"); strcpy(app.name, "test app"); av.init(); av.avg_ncpus = 1; strcpy(result.name, "test result"); result.avp = &av; result.wup = &wu; result.project = &project; result.app = &app; at.result = &result; at.wup = &wu; at.app_version = &av; at.max_elapsed_time = 1e6; at.max_disk_usage = 1e14; at.max_mem_usage = 1e14; strcpy(at.slot_dir, "."); #if 1 // test file copy // ASYNC_COPY* ac = new ASYNC_COPY; FILE_INFO fi; retval = ac->init(&at, &fi, "big_file", "./big_file_copy"); if (retval) { exit(1); } while (1) { do_async_file_ops(); if (at.async_copy == NULL) { break; } } fprintf(stderr, "done\n"); exit(0); #endif ats.active_tasks.push_back(&at); unlink("boinc_finish_called"); unlink("boinc_lockfile"); unlink("boinc_temporary_exit"); unlink("stderr.txt"); retval = at.start(true); if (retval) { fprintf(stderr, "start() failed: %s\n", boincerror(retval)); } while (1) { gstate.now = dtime(); at.preempt(REMOVE_NEVER); ats.poll(); boinc_sleep(.1); at.unsuspend(); ats.poll(); boinc_sleep(.2); //at.request_reread_prefs(); } }
// parse a project's app_info.xml (anonymous platform) file // int CLIENT_STATE::parse_app_info(PROJECT* p, FILE* in) { char buf[256], path[MAXPATHLEN]; MIOFILE mf; mf.init_file(in); XML_PARSER xp(&mf); while (!xp.get_tag()) { if (xp.match_tag("app_info")) continue; if (xp.match_tag("/app_info")) { notices.remove_notices(p, REMOVE_APP_INFO_MSG); return 0; } if (xp.match_tag("file_info") || xp.match_tag("file")) { FILE_INFO* fip = new FILE_INFO; if (fip->parse(xp)) { delete fip; continue; } if (!fip->download_urls.empty() || !fip->upload_urls.empty()) { msg_printf(p, MSG_INFO, "Can't specify URLs in app_info.xml" ); delete fip; continue; } if (link_file_info(p, fip)) { delete fip; continue; } // check that the file is actually there // get_pathname(fip, path, sizeof(path)); if (!boinc_file_exists(path)) { safe_strcpy(buf, _("File referenced in app_info.xml does not exist: ") ); strcat(buf, fip->name); msg_printf(p, MSG_USER_ALERT, "%s", buf); delete fip; continue; } fip->status = FILE_PRESENT; fip->anonymous_platform_file = true; file_infos.push_back(fip); continue; } if (xp.match_tag("app")) { APP* app = new APP; if (app->parse(xp)) { delete app; continue; } if (lookup_app(p, app->name)) { delete app; continue; } link_app(p, app); apps.push_back(app); continue; } if (xp.match_tag("app_version")) { APP_VERSION* avp = new APP_VERSION; if (avp->parse(xp)) { delete avp; continue; } if (cc_config.dont_use_vbox && strstr(avp->plan_class, "vbox")) { msg_printf(p, MSG_INFO, "skipping vbox app in app_info.xml; vbox disabled in cc_config.xml" ); delete avp; continue; } if (strlen(avp->platform) == 0) { safe_strcpy(avp->platform, get_primary_platform()); } if (link_app_version(p, avp)) { delete avp; continue; } app_versions.push_back(avp); continue; } if (log_flags.unparsed_xml) { msg_printf(p, MSG_INFO, "Unparsed line in app_info.xml: %s", xp.parsed_tag ); } } return ERR_XML_PARSE; }
int CLIENT_STATE::parse_state_file_aux(const char* fname) { PROJECT *project=NULL; int retval=0; string stemp; FILE* f = fopen(fname, "r"); if (!f) return ERR_FOPEN; MIOFILE mf; XML_PARSER xp(&mf); mf.init_file(f); while (!xp.get_tag()) { if (xp.match_tag("/client_state")) { break; } if (xp.match_tag("client_state")) { continue; } if (xp.match_tag("project")) { PROJECT temp_project; retval = temp_project.parse_state(xp); if (retval) { msg_printf(NULL, MSG_INTERNAL_ERROR, "Can't parse project in state file"); } else { #ifdef SIM project = new PROJECT; *project = temp_project; projects.push_back(project); #else project = lookup_project(temp_project.master_url); if (project) { project->copy_state_fields(temp_project); } else { msg_printf(&temp_project, MSG_INTERNAL_ERROR, "Project %s is in state file but no account file found", temp_project.get_project_name() ); } #endif } continue; } if (xp.match_tag("app")) { APP* app = new APP; retval = app->parse(xp); if (!project) { msg_printf(NULL, MSG_INTERNAL_ERROR, "Application %s outside project in state file", app->name ); delete app; continue; } if (project->anonymous_platform) { delete app; continue; } if (retval) { msg_printf(NULL, MSG_INTERNAL_ERROR, "Can't parse application in state file" ); delete app; continue; } retval = link_app(project, app); if (retval) { msg_printf(project, MSG_INTERNAL_ERROR, "Can't handle application %s in state file", app->name ); delete app; continue; } apps.push_back(app); continue; } if (xp.match_tag("file_info") || xp.match_tag("file")) { FILE_INFO* fip = new FILE_INFO; retval = fip->parse(xp); if (!project) { msg_printf(NULL, MSG_INTERNAL_ERROR, "File info outside project in state file" ); delete fip; continue; } if (retval) { msg_printf(NULL, MSG_INTERNAL_ERROR, "Can't handle file info in state file" ); delete fip; continue; } retval = link_file_info(project, fip); if (project->anonymous_platform && retval == ERR_NOT_UNIQUE) { delete fip; continue; } if (retval) { msg_printf(project, MSG_INTERNAL_ERROR, "Can't handle file info %s in state file", fip->name ); delete fip; continue; } file_infos.push_back(fip); #ifndef SIM // If the file had a failure before, // don't start another file transfer // int failnum; if (fip->had_failure(failnum)) { if (fip->pers_file_xfer) { delete fip->pers_file_xfer; fip->pers_file_xfer = NULL; } } if (fip->pers_file_xfer) { retval = fip->pers_file_xfer->init(fip, fip->pers_file_xfer->is_upload); if (retval) { msg_printf(project, MSG_INTERNAL_ERROR, "Can't initialize file transfer for %s", fip->name ); } retval = pers_file_xfers->insert(fip->pers_file_xfer); if (retval) { msg_printf(project, MSG_INTERNAL_ERROR, "Can't start persistent file transfer for %s", fip->name ); } } #endif continue; } if (xp.match_tag("app_version")) { APP_VERSION* avp = new APP_VERSION; retval = avp->parse(xp); if (!project) { msg_printf(NULL, MSG_INTERNAL_ERROR, "Application version outside project in state file" ); delete avp; continue; } if (project->anonymous_platform) { delete avp; continue; } if (retval) { msg_printf(NULL, MSG_INTERNAL_ERROR, "Can't parse application version in state file" ); delete avp; continue; } if (strlen(avp->platform) == 0) { safe_strcpy(avp->platform, get_primary_platform()); } else { if (!is_supported_platform(avp->platform)) { // if it's a platform we haven't heard of, // must be that the user tried out a 64 bit client // and then reverted to a 32-bit client. // Let's not throw away the app version and its WUs // #ifndef SIM msg_printf(project, MSG_INTERNAL_ERROR, "App version has unsupported platform %s; changing to %s", avp->platform, get_primary_platform() ); #endif safe_strcpy(avp->platform, get_primary_platform()); } } if (avp->missing_coproc) { msg_printf(project, MSG_INFO, "Application uses missing %s GPU", avp->missing_coproc_name ); } retval = link_app_version(project, avp); if (retval) { delete avp; continue; } app_versions.push_back(avp); continue; } if (xp.match_tag("workunit")) { WORKUNIT* wup = new WORKUNIT; retval = wup->parse(xp); if (!project) { msg_printf(NULL, MSG_INTERNAL_ERROR, "Workunit outside project in state file" ); delete wup; continue; } if (retval) { msg_printf(NULL, MSG_INTERNAL_ERROR, "Can't parse workunit in state file" ); delete wup; continue; } retval = link_workunit(project, wup); if (retval) { msg_printf(project, MSG_INTERNAL_ERROR, "Can't handle workunit in state file" ); delete wup; continue; } workunits.push_back(wup); continue; } if (xp.match_tag("result")) { RESULT* rp = new RESULT; retval = rp->parse_state(xp); if (!project) { msg_printf(NULL, MSG_INTERNAL_ERROR, "Task %s outside project in state file", rp->name ); delete rp; continue; } if (retval) { msg_printf(NULL, MSG_INTERNAL_ERROR, "Can't parse task in state file" ); delete rp; continue; } retval = link_result(project, rp); if (retval) { msg_printf(project, MSG_INTERNAL_ERROR, "Can't link task %s in state file", rp->name ); delete rp; continue; } // handle transition from old clients which didn't store result.platform; // skip for anon platform if (!project->anonymous_platform) { if (!strlen(rp->platform) || !is_supported_platform(rp->platform)) { safe_strcpy(rp->platform, get_primary_platform()); rp->version_num = latest_version(rp->wup->app, rp->platform); } } rp->avp = lookup_app_version( rp->wup->app, rp->platform, rp->version_num, rp->plan_class ); if (!rp->avp) { msg_printf(project, MSG_INTERNAL_ERROR, "No application found for task: %s %d %s; discarding", rp->platform, rp->version_num, rp->plan_class ); delete rp; continue; } if (rp->avp->missing_coproc) { msg_printf(project, MSG_INFO, "Missing coprocessor for task %s", rp->name ); rp->coproc_missing = true; } rp->wup->version_num = rp->version_num; results.push_back(rp); continue; } if (xp.match_tag("project_files")) { if (!project) { msg_printf(NULL, MSG_INTERNAL_ERROR, "Project files outside project in state file" ); xp.skip_unexpected(); continue; } parse_project_files(xp, project->project_files); project->link_project_files(); continue; } if (xp.match_tag("host_info")) { #ifdef SIM retval = host_info.parse(xp, false); coprocs = host_info.coprocs; coprocs.bound_counts(); #else retval = host_info.parse(xp, true); #endif if (retval) { msg_printf(NULL, MSG_INTERNAL_ERROR, "Can't parse host info in state file" ); } continue; } if (xp.match_tag("time_stats")) { retval = time_stats.parse(xp); if (retval) { msg_printf(NULL, MSG_INTERNAL_ERROR, "Can't parse time stats in state file" ); } continue; } if (xp.match_tag("net_stats")) { retval = net_stats.parse(xp); if (retval) { msg_printf(NULL, MSG_INTERNAL_ERROR, "Can't parse network stats in state file" ); } continue; } if (xp.match_tag("active_task_set")) { retval = active_tasks.parse(xp); if (retval) { msg_printf(NULL, MSG_INTERNAL_ERROR, "Can't parse active tasks in state file" ); } continue; } if (xp.parse_string("platform_name", statefile_platform_name)) { continue; } if (xp.parse_string("alt_platform", stemp)) { continue; } if (xp.parse_int("user_run_request", retval)) { cpu_run_mode.set(retval, 0); continue; } if (xp.parse_int("user_run_prev_request", retval)) { cpu_run_mode.set_prev(retval); continue; } if (xp.parse_int("user_gpu_request", retval)) { gpu_run_mode.set(retval, 0); continue; } if (xp.parse_int("user_gpu_prev_request", retval)) { gpu_run_mode.set_prev(retval); continue; } if (xp.parse_int("user_network_request", retval)) { network_run_mode.set(retval, 0); continue; } if (xp.parse_int("core_client_major_version", old_major_version)) { continue; } if (xp.parse_int("core_client_minor_version", old_minor_version)) { continue; } if (xp.parse_int("core_client_release", old_release)) { continue; } if (xp.parse_str("language", language, sizeof(language))) { continue; } if (xp.match_tag("proxy_info")) { retval = gui_proxy_info.parse(xp); if (retval) { msg_printf(NULL, MSG_INTERNAL_ERROR, "Can't parse proxy info in state file" ); } continue; } if (xp.parse_str("host_venue", main_host_venue, sizeof(main_host_venue))) { continue; } if (xp.parse_double("new_version_check_time", new_version_check_time)) { continue; } if (xp.parse_double("all_projects_list_check_time", all_projects_list_check_time)) { continue; } if (xp.parse_string("newer_version", newer_version)) { continue; } #ifdef ENABLE_AUTO_UPDATE if (xp.match_tag("auto_update")) { if (!project) { msg_printf(NULL, MSG_INTERNAL_ERROR, "auto update outside project in state file" ); xp.skip_unexpected(); continue; } if (!auto_update.parse(xp) && !auto_update.validate_and_link(project)) { auto_update.present = true; } continue; } #endif if (log_flags.unparsed_xml) { msg_printf(0, MSG_INFO, "[unparsed_xml] state_file: unrecognized: %s", xp.parsed_tag ); } xp.skip_unexpected(); } sort_results(); fclose(f); // if total resource share is zero, set all shares to 1 // if (projects.size()) { unsigned int i; double x=0; for (i=0; i<projects.size(); i++) { x += projects[i]->resource_share; } if (!x) { msg_printf(NULL, MSG_INFO, "All projects have zero resource share; setting to 100" ); for (i=0; i<projects.size(); i++) { projects[i]->resource_share = 100; } } } return 0; }
// handle completed results // int handle_results() { DB_SCHED_RESULT_ITEM_SET result_handler; SCHED_RESULT_ITEM* srip; unsigned int i; int retval; RESULT* rp; if (g_request->results.size() == 0) return 0; // allow projects to limit the # of results handled // (in case of server memory limits) // if (config.report_max && (int)g_request->results.size() > config.report_max ) { g_request->results.resize(config.report_max); } // copy reported results to a separate vector, "result_handler", // initially with only the "name" field present // for (i=0; i<g_request->results.size(); i++) { result_handler.add_result(g_request->results[i].name); } // read results from database into "result_handler". // // Quantities that must be read from the DB are those // where srip (see below) appears as an rval. // These are: id, name, server_state, received_time, hostid, validate_state. // // Quantities that must be written to the DB are those for // which srip appears as an lval. These are: // hostid, teamid, received_time, client_state, cpu_time, exit_status, // app_version_num, claimed_credit, server_state, stderr_out, // xml_doc_out, outcome, validate_state, elapsed_time // retval = result_handler.enumerate(); if (retval) { log_messages.printf(MSG_CRITICAL, "[HOST#%d] Batch query failed\n", g_reply->host.id ); } // loop over results reported by client // // A note about acks: we send an ack for result received if either // 1) there's some problem with it (wrong state, host, not in DB) or // 2) we update it successfully. // In other words, the only time we don't ack a result is when // it looks OK but the update failed. // for (i=0; i<g_request->results.size(); i++) { rp = &g_request->results[i]; retval = result_handler.lookup_result(rp->name, &srip); if (retval) { log_messages.printf(MSG_CRITICAL, "[HOST#%d] [RESULT#? %s] reported result not in DB\n", g_reply->host.id, rp->name ); g_reply->result_acks.push_back(std::string(rp->name)); continue; } if (config.debug_handle_results) { log_messages.printf(MSG_NORMAL, "[handle] [HOST#%d] [RESULT#%u] [WU#%u] got result (DB: server_state=%d outcome=%d client_state=%d validate_state=%d delete_state=%d)\n", g_reply->host.id, srip->id, srip->workunitid, srip->server_state, srip->outcome, srip->client_state, srip->validate_state, srip->file_delete_state ); } // Do various sanity checks. // If one of them fails, set srip->id = 0, // which suppresses the DB update later on // // If result has server_state OVER // if outcome NO_REPLY accept it (it's just late). // else ignore it // if (srip->server_state == RESULT_SERVER_STATE_OVER) { const char *msg = NULL; switch (srip->outcome) { case RESULT_OUTCOME_INIT: // should never happen! msg = "this result was never sent"; break; case RESULT_OUTCOME_SUCCESS: // don't replace a successful result! msg = "result already reported as success"; // Client is reporting a result twice. // That could mean it didn't get the first reply. // That reply may have contained new jobs. // So make sure we resend lost jobs // g_wreq->resend_lost_results = true; break; case RESULT_OUTCOME_COULDNT_SEND: // should never happen! msg = "this result couldn't be sent"; break; case RESULT_OUTCOME_CLIENT_ERROR: // should never happen! msg = "result already reported as error"; break; case RESULT_OUTCOME_CLIENT_DETACHED: case RESULT_OUTCOME_NO_REPLY: // result is late in arriving, but keep it anyhow break; case RESULT_OUTCOME_DIDNT_NEED: // should never happen msg = "this result wasn't sent (not needed)"; break; case RESULT_OUTCOME_VALIDATE_ERROR: // we already passed through the validator, so // don't keep the new result msg = "result already reported, validate error"; break; default: msg = "server logic bug; please alert BOINC developers"; break; } if (msg) { if (config.debug_handle_results) { log_messages.printf(MSG_NORMAL, "[handle][HOST#%d][RESULT#%u][WU#%u] result already over [outcome=%d validate_state=%d]: %s\n", g_reply->host.id, srip->id, srip->workunitid, srip->outcome, srip->validate_state, msg ); } srip->id = 0; g_reply->result_acks.push_back(std::string(rp->name)); continue; } } if (srip->server_state == RESULT_SERVER_STATE_UNSENT) { log_messages.printf(MSG_CRITICAL, "[HOST#%d] [RESULT#%u] [WU#%u] got unexpected result: server state is %d\n", g_reply->host.id, srip->id, srip->workunitid, srip->server_state ); srip->id = 0; g_reply->result_acks.push_back(std::string(rp->name)); continue; } if (srip->received_time) { log_messages.printf(MSG_CRITICAL, "[HOST#%d] [RESULT#%u] [WU#%u] already got result, at %s \n", g_reply->host.id, srip->id, srip->workunitid, time_to_string(srip->received_time) ); srip->id = 0; g_reply->result_acks.push_back(std::string(rp->name)); continue; } if (srip->hostid != g_reply->host.id) { log_messages.printf(MSG_CRITICAL, "[HOST#%d] [RESULT#%u] [WU#%u] got result from wrong host; expected [HOST#%d]\n", g_reply->host.id, srip->id, srip->workunitid, srip->hostid ); DB_HOST result_host; retval = result_host.lookup_id(srip->hostid); if (retval) { log_messages.printf(MSG_CRITICAL, "[RESULT#%u] [WU#%u] Can't lookup [HOST#%d]\n", srip->id, srip->workunitid, srip->hostid ); srip->id = 0; g_reply->result_acks.push_back(std::string(rp->name)); continue; } else if (result_host.userid != g_reply->host.userid) { log_messages.printf(MSG_CRITICAL, "[USER#%d] [HOST#%d] [RESULT#%u] [WU#%u] Not even the same user; expected [USER#%d]\n", g_reply->host.userid, g_reply->host.id, srip->id, srip->workunitid, result_host.userid ); srip->id = 0; g_reply->result_acks.push_back(std::string(rp->name)); continue; } else { log_messages.printf(MSG_CRITICAL, "[HOST#%d] [RESULT#%u] [WU#%u] Allowing result because same USER#%d\n", g_reply->host.id, srip->id, srip->workunitid, g_reply->host.userid ); } } // hostids do not match // Modify the in-memory copy obtained from the DB earlier. // If we found a problem above, // we have continued and skipped this modify // srip->hostid = g_reply->host.id; srip->teamid = g_reply->user.teamid; srip->received_time = time(0); srip->client_state = rp->client_state; srip->cpu_time = rp->cpu_time; srip->elapsed_time = rp->elapsed_time; // Some buggy clients sporadically report very low elapsed time // but actual CPU time. // Try to fix the elapsed time, since it's critical to credit // if (srip->elapsed_time < srip->cpu_time) { int avid = srip->app_version_id; if (avid > 0) { APP_VERSION* avp = ssp->lookup_app_version(avid); if (avp && !avp->is_multithread()) { srip->elapsed_time = srip->cpu_time; } } } // check for impossible elapsed time // if (srip->elapsed_time < 0) { log_messages.printf(MSG_NORMAL, "[HOST#%d] [RESULT#%u] [WU#%u] negative elapsed time: %f\n", srip->hostid, srip->id, srip->workunitid, srip->elapsed_time ); srip->elapsed_time = 0; } double turnaround_time = srip->received_time - srip->sent_time; if (turnaround_time < 0) { log_messages.printf(MSG_CRITICAL, "[HOST#%d] [RESULT#%u] [WU#%u] inconsistent sent/received times\n", srip->hostid, srip->id, srip->workunitid ); } else { if (srip->elapsed_time > turnaround_time) { log_messages.printf(MSG_NORMAL, "[HOST#%d] [RESULT#%u] [WU#%u] impossible elapsed time: reported %f > turnaround %f\n", srip->hostid, srip->id, srip->workunitid, srip->elapsed_time, turnaround_time ); srip->elapsed_time = turnaround_time; } } srip->exit_status = rp->exit_status; srip->app_version_num = rp->app_version_num; srip->server_state = RESULT_SERVER_STATE_OVER; strlcpy(srip->stderr_out, rp->stderr_out, sizeof(srip->stderr_out)); strlcpy(srip->xml_doc_out, rp->xml_doc_out, sizeof(srip->xml_doc_out)); // look for exit status and app version in stderr_out // (historical - can be deleted at some point) // parse_int(srip->stderr_out, "<exit_status>", srip->exit_status); parse_int(srip->stderr_out, "<app_version>", srip->app_version_num); if ((srip->client_state == RESULT_FILES_UPLOADED) && (srip->exit_status == 0)) { srip->outcome = RESULT_OUTCOME_SUCCESS; if (config.debug_handle_results) { log_messages.printf(MSG_NORMAL, "[handle] [RESULT#%u] [WU#%u]: setting outcome SUCCESS\n", srip->id, srip->workunitid ); } got_good_result(*srip); if (config.dont_store_success_stderr) { strcpy(srip->stderr_out, ""); } } else { if (config.debug_handle_results) { log_messages.printf(MSG_NORMAL, "[handle] [RESULT#%u] [WU#%u]: client_state %d exit_status %d; setting outcome ERROR\n", srip->id, srip->workunitid, srip->client_state, srip->exit_status ); } srip->outcome = RESULT_OUTCOME_CLIENT_ERROR; srip->validate_state = VALIDATE_STATE_INVALID; // adjust quota and reset error rate // got_bad_result(*srip); } } // loop over all incoming results // Update the result records // (skip items that we previously marked to skip) // for (i=0; i<result_handler.results.size(); i++) { SCHED_RESULT_ITEM& sri = result_handler.results[i]; if (sri.id == 0) continue; retval = result_handler.update_result(sri); if (retval) { log_messages.printf(MSG_CRITICAL, "[HOST#%d] [RESULT#%u] [WU#%u] can't update result: %s\n", g_reply->host.id, sri.id, sri.workunitid, boinc_db.error_string() ); } else { g_reply->result_acks.push_back(std::string(sri.name)); } } // set transition_time for the results' WUs // retval = result_handler.update_workunits(); if (retval) { log_messages.printf(MSG_CRITICAL, "[HOST#%d] can't update WUs: %s\n", g_reply->host.id, boincerror(retval) ); } return 0; }