boinc/sched/file_deleter.C

623 lines
20 KiB
C

// Berkeley Open Infrastructure for Network Computing
// http://boinc.berkeley.edu
// Copyright (C) 2005 University of California
//
// This is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation;
// either version 2.1 of the License, or (at your option) any later version.
//
// This software is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
// See the GNU Lesser General Public License for more details.
//
// To view the GNU Lesser General Public License visit
// http://www.gnu.org/copyleft/lesser.html
// or write to the Free Software Foundation, Inc.,
// 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
// file_deleter: deletes files that are no longer needed
#include "config.h"
#include <list>
#include <cstring>
#include <cstdlib>
#include <ctime>
#include <unistd.h>
#include <errno.h>
#include <pwd.h>
#include <sys/types.h>
#include <sys/stat.h>
#ifdef HAVE_STRINGS_H
#include <strings.h>
#endif
#include "boinc_db.h"
#include "parse.h"
#include "error_numbers.h"
#include "str_util.h"
#include "filesys.h"
#include "strings.h"
#include "sched_config.h"
#include "sched_util.h"
#include "sched_msgs.h"
using namespace std;
#define LOCKFILE "file_deleter.out"
#define PIDFILE "file_deleter.pid"
#define SLEEP_INTERVAL 5
#define RESULTS_PER_WU 4 // an estimate of redundancy
SCHED_CONFIG config;
int id_modulus=0, id_remainder=0;
// Given a filename, find its full path in the upload directory hierarchy
// Return an error if file isn't there.
//
int get_file_path(
const char *filename, char* upload_dir, int fanout, char* path
) {
dir_hier_path(filename, upload_dir, fanout, path);
if (boinc_file_exists(path)) {
return 0;
}
return ERR_NOT_FOUND;
}
int wu_delete_files(WORKUNIT& wu) {
char* p;
char filename[256], pathname[256], buf[LARGE_BLOB_SIZE];
bool no_delete=false;
int count_deleted = 0, retval, mthd_retval = 0;
if (strstr(wu.name, "nodelete")) return 0;
safe_strcpy(buf, wu.xml_doc);
p = strtok(buf, "\n");
strcpy(filename, "");
while (p) {
if (parse_str(p, "<name>", filename, sizeof(filename))) {
} else if (match_tag(p, "<file_info>")) {
no_delete = false;
strcpy(filename, "");
} else if (match_tag(p, "<no_delete/>")) {
no_delete = true;
} else if (match_tag(p, "</file_info>")) {
if (!no_delete) {
retval = get_file_path(filename, config.download_dir, config.uldl_dir_fanout,
pathname
);
if (retval) {
log_messages.printf(SCHED_MSG_LOG::MSG_CRITICAL,
"[WU#%d] get_file_path: %s: %d\n",
wu.id, filename, retval
);
} else {
log_messages.printf(SCHED_MSG_LOG::MSG_NORMAL,
"[WU#%d] deleting %s\n", wu.id, filename
);
retval = unlink(pathname);
if (retval && strlen(config.download_dir_alt)) {
sprintf(pathname, "%s/%s", config.download_dir_alt, filename);
retval = unlink(pathname);
}
if (retval) {
log_messages.printf(SCHED_MSG_LOG::MSG_CRITICAL,
"[WU#%d] unlink %s failed: %d\n",
wu.id, filename, retval
);
mthd_retval = ERR_UNLINK;
} else {
count_deleted++;
}
}
}
}
p = strtok(0, "\n");
}
log_messages.printf(SCHED_MSG_LOG::MSG_DEBUG,
"[WU#%d] deleted %d file(s)\n", wu.id, count_deleted
);
return mthd_retval;
}
int result_delete_files(RESULT& result) {
char* p;
char filename[256], pathname[256], buf[LARGE_BLOB_SIZE];
bool no_delete=false;
int count_deleted = 0, retval, mthd_retval = 0;
safe_strcpy(buf, result.xml_doc_in);
p = strtok(buf,"\n");
while (p) {
if (parse_str(p, "<name>", filename, sizeof(filename))) {
} else if (match_tag(p, "<file_info>")) {
no_delete = false;
strcpy(filename, "");
} else if (match_tag(p, "<no_delete/>")) {
no_delete = true;
} else if (match_tag(p, "</file_info>")) {
if (!no_delete) {
retval = get_file_path(
filename, config.upload_dir, config.uldl_dir_fanout,
pathname
);
if (retval) {
// the fact that no result files were found is a critical
// error if this was a successful result, but is to be
// expected if the result outcome was failure, since in
// that case there may well be no output file produced.
//
int debug_or_crit;
if (RESULT_OUTCOME_SUCCESS == result.outcome) {
debug_or_crit=SCHED_MSG_LOG::MSG_CRITICAL;
} else {
debug_or_crit=SCHED_MSG_LOG::MSG_DEBUG;
}
log_messages.printf(debug_or_crit,
"[RESULT#%d] outcome=%d client_state=%d No file %s to delete\n",
result.id, result.outcome, result.client_state, filename
);
} else {
retval = unlink(pathname);
if (retval) {
mthd_retval = ERR_UNLINK;
log_messages.printf(SCHED_MSG_LOG::MSG_CRITICAL,
"[RESULT#%d] unlink %s returned %d %s\n",
result.id, pathname, retval,
(retval && errno)?strerror(errno):""
);
} else {
count_deleted++;
log_messages.printf(SCHED_MSG_LOG::MSG_NORMAL,
"[RESULT#%d] unlinked %s\n", result.id, pathname
);
}
}
}
}
p = strtok(0, "\n");
}
log_messages.printf(SCHED_MSG_LOG::MSG_DEBUG,
"[RESULT#%d] deleted %d file(s)\n", result.id, count_deleted
);
return mthd_retval;
}
// set by corresponding command line arguments.
static bool preserve_wu_files=false;
static bool preserve_result_files=false;
// return nonzero if did anything
//
bool do_pass(bool retry_error) {
DB_WORKUNIT wu;
DB_RESULT result;
bool did_something = false;
char buf[256];
char clause[256];
int retval;
int result_loop_count;
check_stop_daemons();
strcpy(clause, "");
if (id_modulus) {
sprintf(clause, " and id %% %d = %d ", id_modulus, id_remainder);
}
if (config.dont_delete_batches) {
strcat(clause, " and batch <= 0 ");
}
if (retry_error) {
sprintf(buf,
"where file_delete_state=%d or file_delete_state=%d %s limit 1000",
FILE_DELETE_READY, FILE_DELETE_ERROR, clause
);
} else {
sprintf(buf,
"where file_delete_state=%d %s limit 1000",
FILE_DELETE_READY, clause
);
}
while (!wu.enumerate(buf)) {
did_something = true;
retval = 0;
if (!preserve_wu_files) {
retval = wu_delete_files(wu);
}
if (retval) {
wu.file_delete_state = FILE_DELETE_ERROR;
log_messages.printf(SCHED_MSG_LOG::MSG_CRITICAL,
"[WU#%d] update failed: %d\n", wu.id, retval
);
} else {
wu.file_delete_state = FILE_DELETE_DONE;
}
sprintf(buf, "file_delete_state=%d", wu.file_delete_state);
retval= wu.update_field(buf);
}
for (result_loop_count=0; result_loop_count < RESULTS_PER_WU; result_loop_count++) {
if (retry_error) {
sprintf(buf,
"where file_delete_state=%d or file_delete_state=%d %s limit 1000",
FILE_DELETE_READY, FILE_DELETE_ERROR, clause);
} else {
sprintf(buf,
"where file_delete_state=%d %s limit 1000",
FILE_DELETE_READY, clause
);
}
while (!result.enumerate(buf)) {
did_something = true;
retval = 0;
if (!preserve_result_files) {
retval = result_delete_files(result);
}
if (retval) {
result.file_delete_state = FILE_DELETE_ERROR;
log_messages.printf(SCHED_MSG_LOG::MSG_CRITICAL,
"[RESULT#%d] update failed: %d\n", result.id, retval
);
} else {
result.file_delete_state = FILE_DELETE_DONE;
}
sprintf(buf, "file_delete_state=%d", result.file_delete_state);
retval= result.update_field(buf);
}
}
return did_something;
}
struct FILE_RECORD {
string name;
int date_modified;
};
bool operator == (const FILE_RECORD& fr1, const FILE_RECORD& fr2) {
return (fr1.date_modified == fr2.date_modified && fr1.name == fr2.name);
}
bool operator < (const FILE_RECORD& fr1, const FILE_RECORD& fr2) {
if (fr1.date_modified < fr2.date_modified)
return true;
else if (fr1.date_modified > fr2.date_modified)
return false;
else if (fr1.name < fr2.name)
return true;
else
return false;
}
// list of antique files to delete. If order is
// NOT sorted by mod time (primary key) and name
// (secondary key) then unsorted==true.
//
std::list<FILE_RECORD> files_to_delete;
bool unsorted = true;
// delete files from list of old files. Returns number
// of files deleted, or negative for error.
int delete_antique_files(int max_to_delete) {
int nfiles=0;
if (files_to_delete.empty()) return 0;
// sort list and eliminate duplicates
//
if (unsorted) {
files_to_delete.sort();
files_to_delete.unique();
unsorted = false;
}
while (nfiles<max_to_delete && !files_to_delete.empty()) {
char timestamp[128];
char pathname[1024];
int retval;
FILE_RECORD fr = files_to_delete.front();
check_stop_daemons();
retval = get_file_path(
fr.name.c_str(), config.upload_dir,
config.uldl_dir_fanout, pathname
);
if (retval) {
log_messages.printf(SCHED_MSG_LOG::MSG_CRITICAL,
"get_file_path(%s) failed: %d\n",
fr.name.c_str(), retval
);
return -1;
}
strcpy(timestamp, time_to_string(fr.date_modified));
log_messages.printf(SCHED_MSG_LOG::MSG_DEBUG,
"deleting [antique %s] %s\n",
timestamp, pathname
);
if (unlink(pathname)) {
int save_error=errno;
log_messages.printf(SCHED_MSG_LOG::MSG_CRITICAL,
"unlink(%s) failed: %s\n",
pathname, strerror(save_error)
);
return -1;
} else {
nfiles++;
files_to_delete.pop_front();
}
} // while
return nfiles;
}
// construct a list of old files. return number of files
// added to list, or negative number for error.
//
int add_antiques_to_list(int days) {
char command[256];
char single_line[1024];
FILE *fp;
int dirlen=strlen(config.upload_dir);
struct passwd *apache_info=getpwnam("apache");
int del_time=time(0)-86400*days;
int nfiles=0;
if (!apache_info) {
log_messages.printf(SCHED_MSG_LOG::MSG_CRITICAL,
"no user named 'apache' found!\n"
);
return -1;
}
sprintf(command, "find %s -type f -mtime +%d -follow", config.upload_dir, days);
// Now execute the command, read output on a stream. We could use
// find to also exec a 'delete' command. But we want to log all
// file names into the log, and do lots of sanity checking, so
// this way is better.
//
if (!(fp=popen(command, "r"))) {
log_messages.printf(SCHED_MSG_LOG::MSG_CRITICAL,
"command %s failed\n", command
);
return -2;
}
while (fgets(single_line, 1024, fp)) {
char pathname[1024];
char *fname_at_end=NULL;
int nchars=strlen(single_line);
struct stat statbuf;
const char *err=NULL;
FILE_RECORD fr;
// We can interrupt this at any point.
// pclose() is called when process exits.
check_stop_daemons();
// Do serious sanity checking on the path before
// adding the file!!
//
if (!err && nchars > 1022) err="line too long";
if (!err && nchars < dirlen + 1) err="path shorter than upload directory name";
if (!err && single_line[nchars-1] != '\n') err="no newline terminator in line";
if (!err && strncmp(config.upload_dir, single_line, dirlen)) err="upload directory not in path";
if (!err && single_line[dirlen] != '/') err="no slash separator in path";
if (!err) single_line[nchars-1]='\0';
if (!err && stat(single_line, &statbuf)) err="stat failed";
if (!err && statbuf.st_mtime > del_time) err="file too recent";
if (!err && apache_info->pw_uid != statbuf.st_uid) err="file not owned by httpd user";
if (!err && !(fname_at_end=rindex(single_line+dirlen, '/'))) err="no trailing filename";
if (!err) fname_at_end++;
if (!err && !strlen(fname_at_end)) err="trailing filename too short";
// skip NFS file system markers of form .nfs*
//
if (!err && !strncmp(fname_at_end, ".nfs", 4)) {
log_messages.printf(SCHED_MSG_LOG::MSG_CRITICAL,
"Ignoring antique (stale) NFS lockfile %s\n", single_line
);
continue;
}
if (!err && get_file_path(fname_at_end, config.upload_dir, config.uldl_dir_fanout, pathname)) err="get_file_path() failed";
if (!err && strcmp(pathname, single_line)) err="file in wrong hierarchical upload subdirectory";
if (err) {
log_messages.printf(SCHED_MSG_LOG::MSG_CRITICAL,
"Can't list %s for deletion: %s\n",
single_line, err
);
// This file deleting business is SERIOUS. Give up at the
// first sign of ANYTHING amiss.
//
pclose(fp);
return -3;
}
// insert this file onto the list
fr.date_modified = statbuf.st_mtime;
fr.name = fname_at_end;
files_to_delete.push_back(fr);
unsorted = true;
nfiles++;
} // while (fgets(single_line, 1024, fp)) {
pclose(fp);
log_messages.printf(SCHED_MSG_LOG::MSG_DEBUG,
"Found %d antique files to delete\n",
nfiles
);
return nfiles;
}
// returns number of files found & added, or negative for error.
//
int find_antique_files() {
char buf[256];
DB_WORKUNIT wu;
static int nextrun=0;
check_stop_daemons();
// list files once per day
//
if (time(0) < nextrun) return 0;
nextrun = time(0)+86400;
// Find the oldest workunit. We could add "where
// file_delete_state!=FILE_DELETE_DONE" to the query, but this
// might create some race condition with the 'regular' file delete
// mechanism, so better to do it like this.
//
sprintf(buf, "order by create_time limit 1");
if (!wu.enumerate(buf)) {
// Don't ever delete files younger than a month.
//
int days = 1 + (time(0) - wu.create_time)/86400;
if (days<31) days=31;
return add_antiques_to_list(days);
}
return 0;
}
bool do_antique_pass(bool& delete_antiques) {
bool did_something=false;
if (!delete_antiques) return did_something;
// If any problems appear in deleting antique files
// immediately DISABLE this feature.
//
int howmany = find_antique_files();
if (howmany < 0) {
log_messages.printf(SCHED_MSG_LOG::MSG_CRITICAL,
"Problem 1 [%d] in antique file deletion: turning OFF -delete_antiques switch\n", howmany
);
delete_antiques=false;
return did_something;
}
howmany = delete_antique_files(10);
if (howmany < 0) {
log_messages.printf(SCHED_MSG_LOG::MSG_CRITICAL,
"Problem 2 [%d] in antique file deletion: turning OFF -delete_antiques switch\n", howmany
);
delete_antiques=false;
}
if (howmany > 0) did_something=true;
return did_something;
}
int main(int argc, char** argv) {
int retval;
bool asynch = false, one_pass = false, retry_error = false, delete_antiques = false;
int i;
check_stop_daemons();
for (i=1; i<argc; i++) {
if (!strcmp(argv[i], "-asynch")) {
asynch = true;
} else if (!strcmp(argv[i], "-one_pass")) {
one_pass = true;
} else if (!strcmp(argv[i], "-retry_error")) {
retry_error=true;
} else if (!strcmp(argv[i], "-preserve_wu_files")) {
// This option is primarily for testing.
// If enabled, the file_deleter will function 'normally'
// and will update the database,
// but will not actually delete the workunit input files.
// It's equivalent to setting <no_delete/>
// for all workunit input files.
//
preserve_wu_files = true;
} else if (!strcmp(argv[i], "-preserve_result_files")) {
// This option is primarily for testing.
// If enabled, the file_deleter will function 'normally'
// and will update the database,
// but will not actually delete the result output files.
// It's equivalent to setting <no_delete/>
// for all result output files.
//
preserve_result_files = true;
} else if (!strcmp(argv[i], "-d")) {
log_messages.set_debug_level(atoi(argv[++i]));
} else if (!strcmp(argv[i], "-mod")) {
id_modulus = atoi(argv[++i]);
id_remainder = atoi(argv[++i]);
} else if (!strcmp(argv[i], "-delete_antiques")) {
// If turned on, look for and then delete files in the
// upload/ directory which are OLDER than any workunit in
// the database. Such files are ones that were uploaded
// by hosts which turned in results far after the
// deadline. These files are not deleted from the upload
// directory unless you have this option enabled.
delete_antiques = true;
} else {
log_messages.printf(SCHED_MSG_LOG::MSG_CRITICAL, "Unrecognized arg: %s\n", argv[i]);
}
}
if (id_modulus) {
log_messages.printf(SCHED_MSG_LOG::MSG_DEBUG, "Using mod'ed WU/result enumeration. modulus = %d remainder = %d\n",
id_modulus, id_remainder);
}
retval = config.parse_file("..");
if (retval) {
log_messages.printf(SCHED_MSG_LOG::MSG_CRITICAL, "Can't parse config file\n");
exit(1);
}
if (asynch) {
if (fork()) {
exit(0);
}
}
log_messages.printf(SCHED_MSG_LOG::MSG_NORMAL, "Starting\n");
retval = boinc_db.open(config.db_name, config.db_host, config.db_user, config.db_passwd);
if (retval) {
log_messages.printf(SCHED_MSG_LOG::MSG_CRITICAL, "can't open DB\n");
exit(1);
}
retval = boinc_db.set_isolation_level(READ_UNCOMMITTED);
if (retval) {
log_messages.printf(SCHED_MSG_LOG::MSG_CRITICAL,
"boinc_db.set_isolation_level: %d; %s\n", retval, boinc_db.error_string()
);
}
install_stop_signal_handler();
while (1) {
if (!do_pass(retry_error)) {
if (one_pass) break;
if (!do_antique_pass(delete_antiques)) {
sleep(SLEEP_INTERVAL);
}
}
}
}
const char *BOINC_RCSID_bd0d4938a6 = "$Id$";