1 | /* $Id$ */ |
---|
2 | /* |
---|
3 | * FedStage DRMAA for PBS Pro |
---|
4 | * Copyright (C) 2006-2007 FedStage Systems |
---|
5 | * |
---|
6 | * This program is free software: you can redistribute it and/or modify |
---|
7 | * it under the terms of the GNU General Public License as published by |
---|
8 | * the Free Software Foundation, either version 3 of the License, or |
---|
9 | * (at your option) any later version. |
---|
10 | * |
---|
11 | * This program is distributed in the hope that it will be useful, |
---|
12 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
---|
13 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
---|
14 | * GNU General Public License for more details. |
---|
15 | * |
---|
16 | * You should have received a copy of the GNU General Public License |
---|
17 | * along with this program. If not, see <http://www.gnu.org/licenses/>. |
---|
18 | */ |
---|
19 | |
---|
20 | #ifdef HAVE_CONFIG_H |
---|
21 | # include <config.h> |
---|
22 | #endif |
---|
23 | |
---|
24 | #include <pbs_error.h> |
---|
25 | |
---|
26 | #include <drmaa_utils/datetime.h> |
---|
27 | #include <drmaa_utils/drmaa.h> |
---|
28 | #include <drmaa_utils/iter.h> |
---|
29 | #include <drmaa_utils/conf.h> |
---|
30 | #include <drmaa_utils/session.h> |
---|
31 | #include <drmaa_utils/datetime.h> |
---|
32 | |
---|
33 | #include <pbs_drmaa/pbs_conn.h> |
---|
34 | #include <pbs_drmaa/util.h> |
---|
35 | |
---|
36 | #include <errno.h> |
---|
37 | #include <signal.h> |
---|
38 | #include <unistd.h> |
---|
39 | |
---|
40 | |
---|
41 | static char* pbsdrmaa_pbs_submit( pbsdrmaa_pbs_conn_t *self, struct attropl *attrib, char *script, char *destination ); |
---|
42 | |
---|
43 | static struct batch_status* pbsdrmaa_pbs_statjob( pbsdrmaa_pbs_conn_t *self, char *job_id, struct attrl *attrib ); |
---|
44 | |
---|
45 | static void pbsdrmaa_pbs_statjob_free( pbsdrmaa_pbs_conn_t *self, struct batch_status* job_status ); |
---|
46 | |
---|
47 | static void pbsdrmaa_pbs_sigjob( pbsdrmaa_pbs_conn_t *self, char *job_id, char *signal ); |
---|
48 | |
---|
49 | static void pbsdrmaa_pbs_deljob( pbsdrmaa_pbs_conn_t *self, char *job_id ); |
---|
50 | |
---|
51 | static void pbsdrmaa_pbs_rlsjob( pbsdrmaa_pbs_conn_t *self, char *job_id ); |
---|
52 | |
---|
53 | static void pbsdrmaa_pbs_holdjob( pbsdrmaa_pbs_conn_t *self, char *job_id ); |
---|
54 | |
---|
55 | static void pbsdrmaa_pbs_reconnect_internal( pbsdrmaa_pbs_conn_t *self, bool reconnect); |
---|
56 | |
---|
57 | static void pbsdrmaa_pbs_check_connect_internal( pbsdrmaa_pbs_conn_t *self, bool reconnect); |
---|
58 | |
---|
59 | #define IS_TRANSIENT_ERROR (pbs_errno == PBSE_PROTOCOL || pbs_errno == PBSE_EXPIRED || pbs_errno == PBSOLDE_PROTOCOL || pbs_errno == PBSOLDE_EXPIRED) |
---|
60 | |
---|
61 | |
---|
62 | pbsdrmaa_pbs_conn_t * |
---|
63 | pbsdrmaa_pbs_conn_new( pbsdrmaa_session_t *session, char *server ) |
---|
64 | { |
---|
65 | pbsdrmaa_pbs_conn_t *volatile self = NULL; |
---|
66 | |
---|
67 | fsd_log_enter(("")); |
---|
68 | |
---|
69 | TRY |
---|
70 | { |
---|
71 | fsd_malloc(self, pbsdrmaa_pbs_conn_t ); |
---|
72 | |
---|
73 | self->session = session; |
---|
74 | |
---|
75 | self->submit = pbsdrmaa_pbs_submit; |
---|
76 | self->statjob = pbsdrmaa_pbs_statjob; |
---|
77 | self->statjob_free = pbsdrmaa_pbs_statjob_free; |
---|
78 | self->sigjob = pbsdrmaa_pbs_sigjob; |
---|
79 | self->deljob = pbsdrmaa_pbs_deljob; |
---|
80 | self->rlsjob = pbsdrmaa_pbs_rlsjob; |
---|
81 | self->holdjob = pbsdrmaa_pbs_holdjob; |
---|
82 | |
---|
83 | self->server = fsd_strdup(server); |
---|
84 | |
---|
85 | self->connection_fd = -1; |
---|
86 | self->last_usage = time(NULL); |
---|
87 | |
---|
88 | /*ignore SIGPIPE - otherwise pbs_disconnect cause the program to exit */ |
---|
89 | signal(SIGPIPE, SIG_IGN); |
---|
90 | |
---|
91 | pbsdrmaa_pbs_reconnect_internal(self, false); |
---|
92 | } |
---|
93 | EXCEPT_DEFAULT |
---|
94 | { |
---|
95 | if( self != NULL) |
---|
96 | { |
---|
97 | fsd_free(self->server); |
---|
98 | fsd_free(self); |
---|
99 | |
---|
100 | if (self->connection_fd != -1) |
---|
101 | pbs_disconnect(self->connection_fd); |
---|
102 | } |
---|
103 | |
---|
104 | fsd_exc_reraise(); |
---|
105 | } |
---|
106 | END_TRY |
---|
107 | |
---|
108 | fsd_log_return(("")); |
---|
109 | |
---|
110 | return self; |
---|
111 | } |
---|
112 | |
---|
113 | |
---|
114 | void |
---|
115 | pbsdrmaa_pbs_conn_destroy ( pbsdrmaa_pbs_conn_t * self ) |
---|
116 | { |
---|
117 | fsd_log_enter(("")); |
---|
118 | |
---|
119 | TRY |
---|
120 | { |
---|
121 | if(self != NULL) |
---|
122 | { |
---|
123 | fsd_free(self->server); |
---|
124 | fsd_free(self); |
---|
125 | |
---|
126 | if (self->connection_fd != -1) |
---|
127 | pbs_disconnect(self->connection_fd); |
---|
128 | } |
---|
129 | } |
---|
130 | EXCEPT_DEFAULT |
---|
131 | { |
---|
132 | fsd_exc_reraise(); |
---|
133 | } |
---|
134 | END_TRY |
---|
135 | |
---|
136 | fsd_log_return(("")); |
---|
137 | } |
---|
138 | |
---|
139 | char* |
---|
140 | pbsdrmaa_pbs_submit( pbsdrmaa_pbs_conn_t *self, struct attropl *attrib, char *script, char *destination ) |
---|
141 | { |
---|
142 | char *volatile job_id = NULL; |
---|
143 | volatile bool first_try = true; |
---|
144 | volatile bool conn_lock = false; |
---|
145 | |
---|
146 | fsd_log_enter(("")); |
---|
147 | |
---|
148 | TRY |
---|
149 | { |
---|
150 | conn_lock = fsd_mutex_lock(&self->session->super.drm_connection_mutex); |
---|
151 | |
---|
152 | pbsdrmaa_pbs_reconnect_internal(self, false); |
---|
153 | |
---|
154 | retry: |
---|
155 | job_id = pbs_submit(self->connection_fd, attrib, script, destination, NULL); |
---|
156 | |
---|
157 | fsd_log_info(("pbs_submit(%s, %s) = %s", script, destination, job_id)); |
---|
158 | |
---|
159 | if(job_id == NULL) |
---|
160 | { |
---|
161 | fsd_log_error(( "pbs_submit failed, pbs_errno = %d", pbs_errno )); |
---|
162 | if (IS_TRANSIENT_ERROR && first_try) |
---|
163 | { |
---|
164 | pbsdrmaa_pbs_reconnect_internal(self, true); |
---|
165 | first_try = false; |
---|
166 | goto retry; |
---|
167 | } |
---|
168 | else |
---|
169 | { |
---|
170 | pbsdrmaa_exc_raise_pbs( "pbs_submit"); |
---|
171 | } |
---|
172 | } |
---|
173 | } |
---|
174 | EXCEPT_DEFAULT |
---|
175 | { |
---|
176 | fsd_free(job_id); |
---|
177 | fsd_exc_reraise(); |
---|
178 | } |
---|
179 | FINALLY |
---|
180 | { |
---|
181 | if(conn_lock) |
---|
182 | conn_lock = fsd_mutex_unlock(&self->session->super.drm_connection_mutex); |
---|
183 | } |
---|
184 | END_TRY |
---|
185 | |
---|
186 | |
---|
187 | fsd_log_return(("%s", job_id)); |
---|
188 | |
---|
189 | return job_id; |
---|
190 | } |
---|
191 | |
---|
192 | struct batch_status* |
---|
193 | pbsdrmaa_pbs_statjob( pbsdrmaa_pbs_conn_t *self, char *job_id, struct attrl *attrib ) |
---|
194 | { |
---|
195 | |
---|
196 | } |
---|
197 | |
---|
198 | void |
---|
199 | pbsdrmaa_pbs_statjob_free( pbsdrmaa_pbs_conn_t *self, struct batch_status* job_status ) |
---|
200 | { |
---|
201 | |
---|
202 | |
---|
203 | } |
---|
204 | |
---|
205 | void |
---|
206 | pbsdrmaa_pbs_sigjob( pbsdrmaa_pbs_conn_t *self, char *job_id, char *signal ) |
---|
207 | { |
---|
208 | |
---|
209 | |
---|
210 | } |
---|
211 | |
---|
212 | void |
---|
213 | pbsdrmaa_pbs_deljob( pbsdrmaa_pbs_conn_t *self, char *job_id ) |
---|
214 | { |
---|
215 | |
---|
216 | } |
---|
217 | |
---|
218 | void |
---|
219 | pbsdrmaa_pbs_rlsjob( pbsdrmaa_pbs_conn_t *self, char *job_id ) |
---|
220 | { |
---|
221 | |
---|
222 | |
---|
223 | } |
---|
224 | |
---|
225 | void |
---|
226 | pbsdrmaa_pbs_holdjob( pbsdrmaa_pbs_conn_t *self, char *job_id ) |
---|
227 | { |
---|
228 | |
---|
229 | } |
---|
230 | |
---|
231 | void |
---|
232 | pbsdrmaa_pbs_reconnect_internal( pbsdrmaa_pbs_conn_t *self, bool force_reconnect) |
---|
233 | { |
---|
234 | int tries_left = self->session->max_retries_count; |
---|
235 | int sleep_time = 1; |
---|
236 | |
---|
237 | fsd_log_enter(("(%d)", self->connection_fd)); |
---|
238 | |
---|
239 | if ( self->connection_fd != -1 ) |
---|
240 | { |
---|
241 | if (!force_reconnect) |
---|
242 | { |
---|
243 | fsd_log_return(("(%d)", self->connection_fd)); |
---|
244 | return; |
---|
245 | } |
---|
246 | else |
---|
247 | { |
---|
248 | pbs_disconnect(self->connection_fd); |
---|
249 | self->connection_fd = -1; |
---|
250 | } |
---|
251 | } |
---|
252 | |
---|
253 | retry_connect: /* Life... */ |
---|
254 | self->connection_fd = pbs_connect( self->server ); |
---|
255 | fsd_log_info(( "pbs_connect(%s) =%d", self->server, self->connection_fd )); |
---|
256 | if( self->connection_fd < 0 && tries_left-- ) |
---|
257 | { |
---|
258 | sleep(sleep_time); |
---|
259 | sleep_time *=2; |
---|
260 | goto retry_connect; |
---|
261 | } |
---|
262 | |
---|
263 | if( self->connection_fd < 0 ) |
---|
264 | pbsdrmaa_exc_raise_pbs( "pbs_connect" ); |
---|
265 | |
---|
266 | fsd_log_return(("(%d)", self->connection_fd)); |
---|
267 | } |
---|
268 | |
---|