boinc/sched/db_dump.C

526 lines
14 KiB
C++
Raw Normal View History

// The contents of this file are subject to the BOINC Public License
// Version 1.0 (the "License"); you may not use this file except in
// compliance with the License. You may obtain a copy of the License at
// http://boinc.berkeley.edu/license_1.0.txt
//
// Software distributed under the License is distributed on an "AS IS"
// basis, WITHOUT WARRANTY OF ANY KIND, either express or implied. See the
// License for the specific language governing rights and limitations
// under the License.
//
// The Original Code is the Berkeley Open Infrastructure for Network Computing.
//
// The Initial Developer of the Original Code is the SETI@home project.
// Portions created by the SETI@home project are Copyright (C) 2002
// University of California at Berkeley. All Rights Reserved.
//
// Contributor(s):
//
// db_dump: dump database views in XML format
//
// usage: db_dump [-dir path] [-summary_recs num_recs_per_file]
// [-detail_recs num_recs_per_file] [-gzip] [-zip]
// files:
// NOTE: the goal is to make the full DB available (view *_id files)
// to those who want it,
// while allowing those who want to see only a particular row,
// or the highest-ranked rows, to get this info with a small download
//
//
// tables.xml
// for each table (team, user, host):
// total number of records
// number of records per file for summary files
// number of records per file for detailed files
// team_total_credit_N.xml
// list of teams by decreasing total credit (summary)
// team_expavg_credit_N.xml
// team_id_N.xml
// teams by ID (detailed, including list of user summaries)
// user_total_credit_N.xml
// list of users by total credit (summary)
// user_expavg_credit_N.xml
// user_id_N.xml
// users by ID (detailed, including list of host summaries)
// host_total_credit_N.xml
// hosts by decreasing total credit (summary)
// host_expavg_credit_N.xml
// host_id_N.xml
// hosts by ID (detailed)
// NOTE: for now we're using verbose XML tag names.
// We may change to short tag names to save bandwidth.
#include <stdio.h>
#include <string.h>
#include <stdlib.h>
#include <unistd.h>
#include "boinc_db.h"
#include "util.h"
#include "sched_config.h"
#include "sched_util.h"
#include "parse.h"
#define LOCKFILE "db_dump.out"
#define DEFAULT_NRECS_PER_FILE_SUMMARY 1000
#define DEFAULT_NRECS_PER_FILE_DETAIL 100
int nrecs_per_file_summary;
int nrecs_per_file_detail;
bool zip_files = false;
string zip_cmd;
// class that automatically compresses on close
class ZFILE {
protected:
FILE* f;
string filename;
bool zip_file;
public:
ZFILE(bool zip_file_ = zip_files) : f(0), zip_file(zip_file_) {}
~ZFILE() { close(); }
operator FILE* () { return f; }
bool operator ! () { return !f; }
void open(const char* filename_format, ...)
{
close();
va_list ap;
char filename_buf[256];
va_start(ap, filename_format);
vsprintf(filename_buf, filename_format, ap);
va_end(ap);
filename = filename_buf;
f = fopen(filename.c_str(), "w");
if (!f) {
fprintf(stderr, "db_dump: Couldn't open %s for output\n", filename.c_str());
}
}
void close()
{
if (f) {
fclose(f);
if (zip_file) {
string cmd = zip_cmd + ' ' + filename;
system(cmd.c_str());
}
f = 0;
}
}
};
// class that automatically opens a new file every N records
class NUMBERED_ZFILE : public ZFILE {
const char* filename_format;
int nrec_max;
int nfile, nrec;
public:
NUMBERED_ZFILE(const char* filename_format_, int nrec_max_)
: filename_format(filename_format_), nrec_max(nrec_max_), nfile(0), nrec(0)
{
}
NUMBERED_ZFILE& operator ++() {
if (!f || nrec >= nrec_max) {
open(filename_format, nfile);
++nfile;
nrec = 0;
}
++nrec;
return *this;
}
};
void write_host(HOST& host, FILE* f, bool detail, bool show_user) {
fprintf(f,
"<host>\n"
" <id>%d</id>\n",
host.id
);
if (show_user) {
DB_USER user;
user.lookup_id(host.userid);
if (user.show_hosts) {
fprintf(f,
" <userid>%d</userid>\n",
host.userid
);
}
}
fprintf(f,
" <total_credit>%f</total_credit>\n"
" <expavg_credit>%f</expavg_credit>\n"
" <p_vendor>%s</p_vendor>\n"
" <p_model>%s</p_model>\n"
" <os_name>%s</os_name>\n"
" <os_version>%s</os_version>\n",
host.total_credit,
host.expavg_credit,
host.p_vendor,
host.p_model,
host.os_name,
host.os_version
);
if (detail) {
fprintf(f,
" <create_time>%d</create_time>\n"
" <timezone>%d</timezone>\n"
" <ncpus>%d</ncpus>\n"
" <p_fpops>%f</p_fpops>\n"
" <p_iops>%f</p_iops>\n"
" <p_membw>%f</p_membw>\n"
" <m_nbytes>%f</m_nbytes>\n"
" <m_cache>%f</m_cache>\n"
" <m_swap>%f</m_swap>\n"
" <d_total>%f</d_total>\n"
" <d_free>%f</d_free>\n"
" <n_bwup>%f</n_bwup>\n"
" <n_bwdown>%f</n_bwdown>\n",
host.create_time,
host.timezone,
host.p_ncpus,
host.p_fpops,
host.p_iops,
host.p_membw,
host.m_nbytes,
host.m_cache,
host.m_swap,
host.d_total,
host.d_free,
host.n_bwup,
host.n_bwdown
);
}
fprintf(f,
"</host>\n"
);
}
void write_user(USER& user, FILE* f, bool detail, bool show_team) {
DB_HOST host;
char buf[256];
fprintf(f,
"<user>\n"
" <id>%d</id>\n"
" <name>%s</name>\n"
" <url>%s</url>\n"
" <country>%s</country>\n"
" <create_time>%d</create_time>\n"
" <total_credit>%f</total_credit>\n"
" <expavg_credit>%f</expavg_credit>\n",
user.id,
user.name,
user.url,
user.country,
user.create_time,
user.total_credit,
user.expavg_credit
);
if (show_team) {
fprintf(f,
" <teamid>%d</teamid>\n",
user.teamid
);
}
if (detail && user.show_hosts) {
sprintf(buf, "where userid=%d", user.id);
while (!host.enumerate(buf)) {
write_host(host, f, false, false);
}
}
fprintf(f,
"</user>\n"
);
}
void write_team(TEAM& team, FILE* f, bool detail) {
DB_USER user;
char buf[MAX_BLOB_SIZE*2];
fprintf(f,
"<team>\n"
" <id>%d</id>\n"
" <name>%s</name>\n"
" <total_credit>%f</total_credit>\n"
" <expavg_credit>%f</expavg_credit>\n"
" <nusers>%d</nusers>\n",
team.id,
team.name,
team.total_credit,
team.expavg_credit,
team.nusers
);
if (detail) {
fprintf(f,
" <create_time>%d</create_time>\n",
team.create_time
);
if (strlen(team.url)) {
fprintf(f,
" <url>%s</url>\n",
team.url
);
}
if (strlen(team.name_html)) {
escape_url(team.name_html, buf);
fprintf(f,
"<name_html>%s</name_html>\n",
buf
);
}
if (strlen(team.description)) {
escape_url(team.description, buf);
fprintf(f,
"<description>%s</description>\n",
buf
);
}
fprintf(f,
" <country>%s</country>\n",
team.country
);
sprintf(buf, "where teamid=%d", team.id);
while (!user.enumerate(buf)) {
write_user(user, f, false, false);
}
}
fprintf(f,
"</team>\n"
);
}
void team_total_credit() {
DB_TEAM team;
NUMBERED_ZFILE f("team_total_credit_%d", nrecs_per_file_summary);
while (!team.enumerate("order by total_credit desc")) {
write_team(team, ++f, false);
}
}
void team_expavg_credit() {
DB_TEAM team;
NUMBERED_ZFILE f("team_expavg_credit_%d", nrecs_per_file_summary);
while (!team.enumerate("order by expavg_credit desc")) {
write_team(team, ++f, false);
}
}
void team_id() {
DB_TEAM team;
NUMBERED_ZFILE f("team_id_%d", nrecs_per_file_detail);
while (!team.enumerate("order by id")) {
write_team(team, ++f, true);
}
}
void user_total_credit() {
DB_USER user;
NUMBERED_ZFILE f("user_total_credit_%d", nrecs_per_file_summary);
while (!user.enumerate("order by total_credit desc")) {
write_user(user, ++f, false, true);
}
}
void user_expavg_credit() {
DB_USER user;
NUMBERED_ZFILE f("user_expavg_credit_%d", nrecs_per_file_summary);
while (!user.enumerate("order by expavg_credit desc")) {
write_user(user, ++f, false, true);
}
}
void user_id() {
DB_USER user;
NUMBERED_ZFILE f("user_id_%d", nrecs_per_file_detail);
while (!user.enumerate("order by id")) {
write_user(user, ++f, true, true);
}
}
void host_total_credit() {
DB_HOST host;
NUMBERED_ZFILE f("host_total_credit_%d", nrecs_per_file_summary);
while (!host.enumerate("order by total_credit desc")) {
write_host(host, ++f, false, true);
}
}
void host_expavg_credit() {
DB_HOST host;
NUMBERED_ZFILE f("host_expavg_credit_%d", nrecs_per_file_summary);
while (!host.enumerate("order by expavg_credit desc")) {
write_host(host, ++f, false, true);
}
}
void host_id() {
DB_HOST host;
NUMBERED_ZFILE f("host_id_%d", nrecs_per_file_detail);
while (!host.enumerate("order by id")) {
write_host(host, ++f, true, true);
}
}
void core_versions() {
ZFILE f(false);
f.open("core_versions.xml");
if (!f) return;
fprintf(f, "<core_versions>\n");
DB_PLATFORM platform;
while (!platform.enumerate("order by name")) {
DB_CORE_VERSION core_version;
char query_buf[256];
sprintf(query_buf, "where platformid=%d order by version_num desc", platform.id);
if (!core_version.enumerate(query_buf)) {
char url[256] = "";
parse_str(core_version.xml_doc, "<url>", url, sizeof(url));
fprintf(f,
" <core_version>\n"
" <id>%d</id>\n"
" <platform id=\"%d\" name=\"%s\">%s</platform>\n"
" <version>%d</version>\n"
" <create_time>%d</create_time>\n"
" <url>%s</url>\n"
" </core_version>\n",
core_version.id,
platform.id, platform.name, platform.user_friendly_name,
core_version.version_num,
core_version.create_time,
url);
}
}
fprintf(f, "</core_versions>\n");
}
int tables_file() {
int nusers, nteams, nhosts;
int retval;
DB_USER user;
DB_TEAM team;
DB_HOST host;
ZFILE f(false);
f.open("tables.xml");
if (!f) return -1;
retval = user.count(nusers);
if (retval) return retval;
retval = team.count(nteams);
if (retval) return retval;
retval = host.count(nhosts);
if (retval) return retval;
fprintf(f,
"<tables>\n"
" <update_time>%d</update_time>\n"
" <nusers_total>%d</nusers_total>\n"
" <nusers_per_file_summary>%d</nusers_per_file_summary>\n"
" <nusers_per_file_detail>%d</nusers_per_file_detail>\n"
" <nteams_total>%d</nteams_total>\n"
" <nteams_per_file_summary>%d</nteams_per_file_summary>\n"
" <nteams_per_file_detail>%d</nteams_per_file_detail>\n"
" <nhosts_total>%d</nhosts_total>\n"
" <nhosts_per_file_summary>%d</nhosts_per_file_summary>\n"
" <nhosts_per_file_detail>%d</nhosts_per_file_detail>\n"
"</tables>\n",
(int)time(0),
nusers,
nrecs_per_file_summary,
nrecs_per_file_detail,
nteams,
nrecs_per_file_summary,
nrecs_per_file_detail,
nhosts,
nrecs_per_file_summary,
nrecs_per_file_detail
);
return 0;
}
int main(int argc, char** argv) {
SCHED_CONFIG config;
int retval, i;
char dir[256];
check_stop_trigger();
nrecs_per_file_summary = DEFAULT_NRECS_PER_FILE_SUMMARY;
nrecs_per_file_detail = DEFAULT_NRECS_PER_FILE_DETAIL;
strcpy(dir, "");
for (i=1; i<argc; i++) {
if (!strcmp(argv[i], "-dir")) {
strcpy(dir, argv[++i]);
} else if (!strcmp(argv[i], "-d")) {
log_messages.set_debug_level(atoi(argv[++i]));
} else if (!strcmp(argv[i], "-gzip")) {
zip_files = true;
zip_cmd = "gzip -fq";
} else if (!strcmp(argv[i], "-zip")) {
zip_files = true;
zip_cmd = "zip -q";
} else if (!strcmp(argv[i], "-summary_recs")) {
nrecs_per_file_summary = atoi(argv[++i]);
} else if (!strcmp(argv[i], "-detail_recs")) {
nrecs_per_file_detail = atoi(argv[++i]);
}
}
if (nrecs_per_file_summary <= 0 || nrecs_per_file_detail <= 0) {
log_messages.printf(SchedMessages::NORMAL, "Too few records per file.\n");
exit(1);
}
if (lock_file(LOCKFILE)) {
log_messages.printf(SchedMessages::NORMAL, "Another copy of db_dump is already running\n");
exit(1);
}
log_messages.printf(SchedMessages::NORMAL, "Starting\n");
retval = config.parse_file("..");
if (retval) {
log_messages.printf(SchedMessages::NORMAL, "Can't parse config file\n");
exit(1);
}
retval = boinc_db.open(config.db_name, config.db_passwd);
if (retval) {
log_messages.printf(SchedMessages::NORMAL, "Can't open DB\n");
exit(1);
}
if (strlen(dir)) {
retval = chdir(dir);
if (retval) {
log_messages.printf(SchedMessages::NORMAL, "can't chdir to %s\n", dir);
exit(1);
}
}
tables_file();
team_total_credit();
team_expavg_credit();
team_id();
user_total_credit();
user_expavg_credit();
user_id();
host_total_credit();
host_expavg_credit();
host_id();
core_versions();
}