contrib/check_cluster.c

   1 /*****************************************************************************
   2  *
   3  * CHECK_CLUSTER.C - Host and Service Cluster Plugin for NetSaint
   4  *
   5  * Copyright (c) 2000 Ethan Galstad (netsaint@netsaint.org)
   6  * License: GPL
   7  * Last Modified:   07-08-2000
   8  *
   9  * License:
  10  *
  11  * This program is free software; you can redistribute it and/or modify
  12  * it under the terms of the GNU General Public License as published by
  13  * the Free Software Foundation; either version 2 of the License, or
  14  * (at your option) any later version.
  15  *
  16  * This program is distributed in the hope that it will be useful,
  17  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  18  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  19  * GNU General Public License for more details.
  20  *
  21  * You should have received a copy of the GNU General Public License
  22  * along with this program; if not, write to the Free Software
  23  * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
  24  *
  25  *****************************************************************************/
  26
  27
  28 #include <stdio.h>
  29 #include <stdlib.h>
  30
  31 #define OK              0
  32 #define ERROR           -1
  33
  34 #define TRUE            1
  35 #define FALSE           0
  36
  37 #define CHECK_SERVICES  1
  38 #define CHECK_HOSTS     2
  39
  40 #define MAX_INPUT_BUFFER        1024
  41
  42 #define STATE_OK        0
  43 #define STATE_WARNING   1
  44 #define STATE_CRITICAL  2
  45 #define STATE_UNKNOWN   3
  46
  47 typedef struct clustermember_struct{
  48         char *host_name;
  49         char *svc_description;
  50         struct clustermember_struct *next;
  51         }clustermember;
  52
  53
  54 int check_cluster_status(void);
  55 int add_clustermember(char *,char *);
  56 void free_memory(void);
  57
  58 clustermember *clustermember_list=NULL;
  59
  60 int total_services_ok=0;
  61 int total_services_warning=0;
  62 int total_services_unknown=0;
  63 int total_services_critical=0;
  64
  65 int total_hosts_up=0;
  66 int total_hosts_down=0;
  67 int total_hosts_unreachable=0;
  68
  69 char status_log[MAX_INPUT_BUFFER]="";
  70 int warning_threshold=0;
  71 int critical_threshold=0;
  72
  73 int check_type=CHECK_SERVICES;
  74
  75
  76 int main(int argc, char **argv){
  77         char input_buffer[MAX_INPUT_BUFFER];
  78         char *host_name;
  79         char *svc_description;
  80         int return_code=STATE_OK;
  81         int error=FALSE;
  82
  83         if(argc!=5){
  84
  85                 printf("Invalid arguments supplied\n");
  86                 printf("\n");
  87
  88                 printf("Host/Service Cluster Plugin for NetSaint\n");
  89                 printf("Copyright (c) 2000 Ethan Galstad (netsaint@netsaint.org)\n");
  90                 printf("Last Modified: 07-08-2000\n");
  91                 printf("License: GPL\n");
  92                 printf("\n");
  93                 printf("Usage: %s <--service | --host> <status_log> <warn_threshold> <crit_threshold>\n",argv[0]);
  94                 printf("\n");
  95                 printf("Options:\n");
  96                 printf("   --service        = Check service cluster status\n");
  97                 printf("   --host           = Check host cluster status\n");
  98                 printf("   <status_log>     = This is the location of the NetSaint status log\n");
  99                 printf("   <warn_threshold> = This is the number of hosts or services in\n");
 100                 printf("                      the cluster that must be in a non-OK state\n");
 101                 printf("                      in order to result in a warning status level\n");
 102                 printf("   <crit_threshold> = This is the number of hosts or services in\n");
 103                 printf("                      the cluster that must be in a non-OK state\n");
 104                 printf("                      in order to result in a critical status level\n");
 105                 printf("\n");
 106                 printf("Notes:\n");
 107                 printf("Members of the host or service cluster are read from STDIN.\n");
 108                 printf("One host or service can be specified per line, services must\n");
 109                 printf("be in the format of <host_name>;<svc_description>\n");
 110                 printf("\n");
 111
 112                 return STATE_UNKNOWN;
 113                 }
 114
 115         /* see if we're checking a host or service clust */
 116         if(!strcmp(argv[1],"--host"))
 117                 check_type=CHECK_HOSTS;
 118         else
 119                 check_type=CHECK_SERVICES;
 120
 121         /* get the status log */
 122         strncpy(status_log,argv[2],sizeof(status_log)-1);
 123         status_log[sizeof(status_log)-1]='\x0';
 124
 125         /* get the warning and critical thresholds */
 126         warning_threshold=atoi(argv[3]);
 127         critical_threshold=atoi(argv[4]);
 128
 129
 130         /* read all data from STDIN until there isn't anymore */
 131         while(fgets(input_buffer,sizeof(input_buffer)-1,stdin)){
 132
 133                 if(feof(stdin))
 134                         break;
 135
 136                 /*strip(input_buffer);*/
 137
 138                 if(!strcmp(input_buffer,""))
 139                         continue;
 140
 141                 if(!strcmp(input_buffer,"\n"))
 142                         continue;
 143
 144                 /* get the host name */
 145                 if(check_type==CHECK_SERVICES)
 146                         host_name=(char *)strtok(input_buffer,";");
 147                 else
 148                         host_name=(char *)strtok(input_buffer,"\n");
 149                 if(host_name==NULL || !strcmp(host_name,"")){
 150                         printf("Error: Host name is NULL!\n");
 151                         continue;
 152                         }
 153
 154                 if(check_type==CHECK_SERVICES){
 155
 156                         /* get the service description */
 157                         svc_description=(char *)strtok(NULL,"\n");
 158                         if(svc_description==NULL || !strcmp(svc_description,"")){
 159                                 printf("Error: Service description is NULL!\n");
 160                                 continue;
 161                                 }
 162                         }
 163
 164                 /* add the cluster member to the list in memory */
 165                 if(add_clustermember(host_name,svc_description)!=OK)
 166                         printf("Error: Could not add cluster member\n");
 167 #ifdef DEBUG
 168                 else
 169                         printf("Added cluster member\n");
 170 #endif
 171                 }
 172
 173
 174         /* check the status of the cluster */
 175         if(check_cluster_status()==OK){
 176
 177                 if(check_type==CHECK_SERVICES){
 178                         if((total_services_warning+total_services_unknown+total_services_critical) >= critical_threshold)
 179                                 return_code=STATE_CRITICAL;
 180                         else if((total_services_warning+total_services_unknown+total_services_critical) >= warning_threshold)
 181                                 return_code=STATE_WARNING;
 182                         else
 183                                 return_code=STATE_OK;
 184
 185                         printf("Service cluster %s: %d ok, %d warning, %d unknown, %d critical\n",(return_code==STATE_OK)?"ok":"problem",total_services_ok,total_services_warning,total_services_unknown,total_services_critical);
 186                         }
 187                 else{
 188                         if((total_hosts_down+total_hosts_unreachable) >= critical_threshold)
 189                                 return_code=STATE_CRITICAL;
 190                         else if((total_hosts_down+total_hosts_unreachable) >= warning_threshold)
 191                                 return_code=STATE_WARNING;
 192                         else
 193                                 return_code=STATE_OK;
 194
 195                         printf("Host cluster %s: %d up, %d down, %d unreachable\n",(return_code==STATE_OK)?"ok":"problem",total_hosts_up,total_hosts_down,total_hosts_unreachable);
 196                         }
 197                 }
 198         else
 199                 return_code=STATE_UNKNOWN;
 200
 201         free_memory();
 202
 203         return return_code;
 204         }
 205
 206
 207
 208 int add_clustermember(char *hst,char *svc){
 209         clustermember *new_clustermember;
 210
 211         new_clustermember=(clustermember *)malloc(sizeof(clustermember));
 212         if(new_clustermember==NULL)
 213                 return ERROR;
 214
 215         new_clustermember->host_name=NULL;
 216         new_clustermember->svc_description=NULL;
 217
 218         if(hst!=NULL){
 219                 new_clustermember->host_name=(char *)malloc(strlen(hst)+1);
 220                 if(new_clustermember->host_name==NULL){
 221                         free(new_clustermember);
 222                         return ERROR;
 223                         }
 224                 strcpy(new_clustermember->host_name,hst);
 225                 }
 226
 227         if(svc!=NULL){
 228                 new_clustermember->svc_description=(char *)malloc(strlen(svc)+1);
 229                 if(new_clustermember->svc_description==NULL){
 230                         if(new_clustermember->host_name!=NULL)
 231                                 free(new_clustermember->host_name);
 232                         free(new_clustermember);
 233                         return ERROR;
 234                         }
 235                 strcpy(new_clustermember->svc_description,svc);
 236                 }
 237
 238         new_clustermember->next=clustermember_list;
 239         clustermember_list=new_clustermember;
 240
 241         return OK;
 242         }
 243
 244
 245 void free_memory(void){
 246         clustermember *this_clustermember;
 247         clustermember *next_clustermember;
 248
 249         for(this_clustermember=clustermember_list;this_clustermember!=NULL;this_clustermember=next_clustermember){
 250                 next_clustermember=this_clustermember->next;
 251                 if(this_clustermember->host_name!=NULL)
 252                         free(this_clustermember->host_name);
 253                 if(this_clustermember->svc_description!=NULL)
 254                         free(this_clustermember->svc_description);
 255                 free(this_clustermember);
 256                 }
 257
 258         return;
 259         }
 260
 261
 262
 263 int check_cluster_status(void){
 264         FILE *fp;
 265         clustermember *temp_clustermember;
 266         char input_buffer[MAX_INPUT_BUFFER];
 267         char matching_entry[MAX_INPUT_BUFFER];
 268
 269         fp=fopen(status_log,"r");
 270         if(fp==NULL){
 271                 printf("Error: Could not open status log '%s' for reading\n",status_log);
 272                 return ERROR;
 273                 }
 274
 275 #ifdef DEBUG
 276         for(temp_clustermember=clustermember_list;temp_clustermember!=NULL;temp_clustermember=temp_clustermember->next){
 277                 if(check_type==CHECK_HOSTS)
 278                         printf("Cluster member: '%s'\n",temp_clustermember->host_name);
 279                 else
 280                         printf("Cluster member: '%s'/'%s'\n",temp_clustermember->host_name,temp_clustermember->svc_description);
 281                 }
 282 #endif
 283
 284         for(fgets(input_buffer,MAX_INPUT_BUFFER-1,fp);!feof(fp);fgets(input_buffer,MAX_INPUT_BUFFER-1,fp)){
 285
 286                 /* this is a host entry */
 287                 if(strstr(input_buffer,"] HOST;") && check_type==CHECK_HOSTS){
 288
 289                         /* this this a match? */
 290                         for(temp_clustermember=clustermember_list;temp_clustermember!=NULL;temp_clustermember=temp_clustermember->next){
 291
 292                                 snprintf(matching_entry,sizeof(matching_entry)-1,";%s;",temp_clustermember->host_name);
 293
 294                                 if(strstr(input_buffer,matching_entry)){
 295                                         if(strstr(input_buffer,";DOWN;"))
 296                                                 total_hosts_down++;
 297                                         else if(strstr(input_buffer,";UNREACHABLE;"))
 298                                                 total_hosts_unreachable++;
 299                                         else if(strstr(input_buffer,";UP;"))
 300                                                 total_hosts_up++;
 301                                         }
 302                                 }
 303
 304                         }
 305
 306                 /* this is a service entry */
 307                 else if(strstr(input_buffer,"] SERVICE;") && check_type==CHECK_SERVICES){
 308
 309                         /* this this a match? */
 310                         for(temp_clustermember=clustermember_list;temp_clustermember!=NULL;temp_clustermember=temp_clustermember->next){
 311
 312                                 snprintf(matching_entry,sizeof(matching_entry)-1,";%s;%s;",temp_clustermember->host_name,temp_clustermember->svc_description);
 313
 314                                 if(strstr(input_buffer,matching_entry)){
 315                                         if(strstr(input_buffer,";HOST DOWN;") || strstr(input_buffer,";UNREACHABLE;") || strstr(input_buffer,";CRITICAL;"))
 316                                                 total_services_critical++;
 317                                         else if(strstr(input_buffer,";WARNING;"))
 318                                                 total_services_warning++;
 319                                         else if(strstr(input_buffer,";UNKNOWN;"))
 320                                                 total_services_unknown++;
 321                                         else if(strstr(input_buffer,";OK;") || strstr(input_buffer,";RECOVERY;"))
 322                                                 total_services_ok++;
 323                                         }
 324                                 }
 325
 326                         }
 327                 }
 328
 329         fclose(fp);
 330
 331         return OK;
 332         }