1 /******************************************************************************
3 * Nagios check_ntp plugin
6 * Copyright (c) 2006 sean finney <seanius@seanius.net>
7 * Copyright (c) 2006 nagios-plugins team
9 * Last Modified: $Date$
13 * This file contains the check_ntp plugin
15 * This plugin to check ntp servers independant of any commandline
16 * programs or external libraries.
19 * License Information:
21 * This program is free software; you can redistribute it and/or modify
22 * it under the terms of the GNU General Public License as published by
23 * the Free Software Foundation; either version 2 of the License, or
24 * (at your option) any later version.
26 * This program is distributed in the hope that it will be useful,
27 * but WITHOUT ANY WARRANTY; without even the implied warranty of
28 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
29 * GNU General Public License for more details.
31 * You should have received a copy of the GNU General Public License
32 * along with this program; if not, write to the Free Software
33 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
37 *****************************************************************************/
39 const char *progname
= "check_ntp";
40 const char *revision
= "$Revision$";
41 const char *copyright
= "2006";
42 const char *email
= "nagiosplug-devel@lists.sourceforge.net";
48 static char *server_address
=NULL
;
50 static double owarn
=60;
51 static double ocrit
=120;
52 static short do_jitter
=0;
53 static double jwarn
=5000;
54 static double jcrit
=10000;
56 int process_arguments (int, char **);
57 void print_help (void);
58 void print_usage (void);
60 /* number of times to perform each request to get a good average. */
63 /* max size of control message data */
64 #define MAX_CM_SIZE 468
66 /* this structure holds everything in an ntp request/response as per rfc1305 */
68 uint8_t flags
; /* byte with leapindicator,vers,mode. see macros */
69 uint8_t stratum
; /* clock stratum */
70 int8_t poll
; /* polling interval */
71 int8_t precision
; /* precision of the local clock */
72 int32_t rtdelay
; /* total rt delay, as a fixed point num. see macros */
73 uint32_t rtdisp
; /* like above, but for max err to primary src */
74 uint32_t refid
; /* ref clock identifier */
75 uint64_t refts
; /* reference timestamp. local time local clock */
76 uint64_t origts
; /* time at which request departed client */
77 uint64_t rxts
; /* time at which request arrived at server */
78 uint64_t txts
; /* time at which request departed server */
81 /* this structure holds data about results from querying offset from a peer */
83 time_t waiting
; /* ts set when we started waiting for a response */
84 int num_responses
; /* number of successfully recieved responses */
85 uint8_t stratum
; /* copied verbatim from the ntp_message */
86 double rtdelay
; /* converted from the ntp_message */
87 double rtdisp
; /* converted from the ntp_message */
88 double offset
[AVG_NUM
]; /* offsets from each response */
89 uint8_t flags
; /* byte with leapindicator,vers,mode. see macros */
92 /* this structure holds everything in an ntp control message as per rfc1305 */
94 uint8_t flags
; /* byte with leapindicator,vers,mode. see macros */
95 uint8_t op
; /* R,E,M bits and Opcode */
96 uint16_t seq
; /* Packet sequence */
97 uint16_t status
; /* Clock status */
98 uint16_t assoc
; /* Association */
99 uint16_t offset
; /* Similar to TCP sequence # */
100 uint16_t count
; /* # bytes of data */
101 char data
[MAX_CM_SIZE
]; /* ASCII data of the request */
102 /* NB: not necessarily NULL terminated! */
103 } ntp_control_message
;
105 /* this is an association/status-word pair found in control packet reponses */
109 } ntp_assoc_status_pair
;
111 /* bits 1,2 are the leap indicator */
113 #define LI(x) ((x&LI_MASK)>>6)
114 #define LI_SET(x,y) do{ x |= ((y<<6)&LI_MASK); }while(0)
115 /* and these are the values of the leap indicator */
116 #define LI_NOWARNING 0x00
117 #define LI_EXTRASEC 0x01
118 #define LI_MISSINGSEC 0x02
119 #define LI_ALARM 0x03
120 /* bits 3,4,5 are the ntp version */
122 #define VN(x) ((x&VN_MASK)>>3)
123 #define VN_SET(x,y) do{ x |= ((y<<3)&VN_MASK); }while(0)
124 #define VN_RESERVED 0x02
125 /* bits 6,7,8 are the ntp mode */
126 #define MODE_MASK 0x07
127 #define MODE(x) (x&MODE_MASK)
128 #define MODE_SET(x,y) do{ x |= (y&MODE_MASK); }while(0)
129 /* here are some values */
130 #define MODE_CLIENT 0x03
131 #define MODE_CONTROLMSG 0x06
132 /* In control message, bits 8-10 are R,E,M bits */
133 #define REM_MASK 0xe0
134 #define REM_RESP 0x80
135 #define REM_ERROR 0x40
136 #define REM_MORE 0x20
137 /* In control message, bits 11 - 15 are opcode */
139 #define OP_SET(x,y) do{ x |= (y&OP_MASK); }while(0)
140 #define OP_READSTAT 0x01
141 #define OP_READVAR 0x02
142 /* In peer status bytes, bits 6,7,8 determine clock selection status */
143 #define PEER_SEL(x) ((ntohs(x)>>8)&0x07)
144 #define PEER_INCLUDED 0x04
145 #define PEER_SYNCSOURCE 0x06
148 ** a note about the 32-bit "fixed point" numbers:
150 they are divided into halves, each being a 16-bit int in network byte order:
151 - the first 16 bits are an int on the left side of a decimal point.
152 - the second 16 bits represent a fraction n/(2^16)
153 likewise for the 64-bit "fixed point" numbers with everything doubled :)
156 /* macros to access the left/right 16 bits of a 32-bit ntp "fixed point"
157 number. note that these can be used as lvalues too */
158 #define L16(x) (((uint16_t*)&x)[0])
159 #define R16(x) (((uint16_t*)&x)[1])
160 /* macros to access the left/right 32 bits of a 64-bit ntp "fixed point"
161 number. these too can be used as lvalues */
162 #define L32(x) (((uint32_t*)&x)[0])
163 #define R32(x) (((uint32_t*)&x)[1])
165 /* ntp wants seconds since 1/1/00, epoch is 1/1/70. this is the difference */
166 #define EPOCHDIFF 0x83aa7e80UL
168 /* extract a 32-bit ntp fixed point number into a double */
169 #define NTP32asDOUBLE(x) (ntohs(L16(x)) + (double)ntohs(R16(x))/65536.0)
171 /* likewise for a 64-bit ntp fp number */
172 #define NTP64asDOUBLE(n) (double)(((uint64_t)n)?\
173 (ntohl(L32(n))-EPOCHDIFF) + \
174 (.00000001*(0.5+(double)(ntohl(R32(n))/42.94967296))):\
177 /* convert a struct timeval to a double */
178 #define TVasDOUBLE(x) (double)(x.tv_sec+(0.000001*x.tv_usec))
180 /* convert an ntp 64-bit fp number to a struct timeval */
181 #define NTP64toTV(n,t) \
182 do{ if(!n) t.tv_sec = t.tv_usec = 0; \
184 t.tv_sec=ntohl(L32(n))-EPOCHDIFF; \
185 t.tv_usec=(int)(0.5+(double)(ntohl(R32(n))/4294.967296)); \
189 /* convert a struct timeval to an ntp 64-bit fp number */
190 #define TVtoNTP64(t,n) \
191 do{ if(!t.tv_usec && !t.tv_sec) n=0x0UL; \
193 L32(n)=htonl(t.tv_sec + EPOCHDIFF); \
194 R32(n)=htonl((uint64_t)((4294.967296*t.tv_usec)+.5)); \
198 /* NTP control message header is 12 bytes, plus any data in the data
199 * field, plus null padding to the nearest 32-bit boundary per rfc.
201 #define SIZEOF_NTPCM(m) (12+ntohs(m.count)+((m.count)?4-(ntohs(m.count)%4):0))
203 /* finally, a little helper or two for debugging: */
204 #define DBG(x) do{if(verbose>1){ x; }}while(0);
205 #define PRINTSOCKADDR(x) \
207 printf("%u.%u.%u.%u", (x>>24)&0xff, (x>>16)&0xff, (x>>8)&0xff, x&0xff);\
210 /* calculate the offset of the local clock */
211 static inline double calc_offset(const ntp_message
*m
, const struct timeval
*t
){
212 double client_tx
, peer_rx
, peer_tx
, client_rx
;
213 client_tx
= NTP64asDOUBLE(m
->origts
);
214 peer_rx
= NTP64asDOUBLE(m
->rxts
);
215 peer_tx
= NTP64asDOUBLE(m
->txts
);
216 client_rx
=TVasDOUBLE((*t
));
217 return (.5*((peer_tx
-client_rx
)+(peer_rx
-client_tx
)));
220 /* print out a ntp packet in human readable/debuggable format */
221 void print_ntp_message(const ntp_message
*p
){
222 struct timeval ref
, orig
, rx
, tx
;
224 NTP64toTV(p
->refts
,ref
);
225 NTP64toTV(p
->origts
,orig
);
226 NTP64toTV(p
->rxts
,rx
);
227 NTP64toTV(p
->txts
,tx
);
229 printf("packet contents:\n");
230 printf("\tflags: 0x%.2x\n", p
->flags
);
231 printf("\t li=%d (0x%.2x)\n", LI(p
->flags
), p
->flags
&LI_MASK
);
232 printf("\t vn=%d (0x%.2x)\n", VN(p
->flags
), p
->flags
&VN_MASK
);
233 printf("\t mode=%d (0x%.2x)\n", MODE(p
->flags
), p
->flags
&MODE_MASK
);
234 printf("\tstratum = %d\n", p
->stratum
);
235 printf("\tpoll = %g\n", pow(2, p
->poll
));
236 printf("\tprecision = %g\n", pow(2, p
->precision
));
237 printf("\trtdelay = %-.16g\n", NTP32asDOUBLE(p
->rtdelay
));
238 printf("\trtdisp = %-.16g\n", NTP32asDOUBLE(p
->rtdisp
));
239 printf("\trefid = %x\n", p
->refid
);
240 printf("\trefts = %-.16g\n", NTP64asDOUBLE(p
->refts
));
241 printf("\torigts = %-.16g\n", NTP64asDOUBLE(p
->origts
));
242 printf("\trxts = %-.16g\n", NTP64asDOUBLE(p
->rxts
));
243 printf("\ttxts = %-.16g\n", NTP64asDOUBLE(p
->txts
));
246 void print_ntp_control_message(const ntp_control_message
*p
){
248 const ntp_assoc_status_pair
*peer
=NULL
;
250 printf("control packet contents:\n");
251 printf("\tflags: 0x%.2x , 0x%.2x\n", p
->flags
, p
->op
);
252 printf("\t li=%d (0x%.2x)\n", LI(p
->flags
), p
->flags
&LI_MASK
);
253 printf("\t vn=%d (0x%.2x)\n", VN(p
->flags
), p
->flags
&VN_MASK
);
254 printf("\t mode=%d (0x%.2x)\n", MODE(p
->flags
), p
->flags
&MODE_MASK
);
255 printf("\t response=%d (0x%.2x)\n", (p
->op
&REM_RESP
)>0, p
->op
&REM_RESP
);
256 printf("\t more=%d (0x%.2x)\n", (p
->op
&REM_MORE
)>0, p
->op
&REM_MORE
);
257 printf("\t error=%d (0x%.2x)\n", (p
->op
&REM_ERROR
)>0, p
->op
&REM_ERROR
);
258 printf("\t op=%d (0x%.2x)\n", p
->op
&OP_MASK
, p
->op
&OP_MASK
);
259 printf("\tsequence: %d (0x%.2x)\n", ntohs(p
->seq
), ntohs(p
->seq
));
260 printf("\tstatus: %d (0x%.2x)\n", ntohs(p
->status
), ntohs(p
->status
));
261 printf("\tassoc: %d (0x%.2x)\n", ntohs(p
->assoc
), ntohs(p
->assoc
));
262 printf("\toffset: %d (0x%.2x)\n", ntohs(p
->offset
), ntohs(p
->offset
));
263 printf("\tcount: %d (0x%.2x)\n", ntohs(p
->count
), ntohs(p
->count
));
264 numpeers
=ntohs(p
->count
)/(sizeof(ntp_assoc_status_pair
));
265 if(p
->op
&REM_RESP
&& p
->op
&OP_READSTAT
){
266 peer
=(ntp_assoc_status_pair
*)p
->data
;
267 for(i
=0;i
<numpeers
;i
++){
268 printf("\tpeer id %.2x status %.2x",
269 ntohs(peer
[i
].assoc
), ntohs(peer
[i
].status
));
270 if (PEER_SEL(peer
[i
].status
) >= PEER_INCLUDED
){
271 if(PEER_SEL(peer
[i
].status
) >= PEER_SYNCSOURCE
){
272 printf(" <-- current sync source");
274 printf(" <-- current sync candidate");
282 void setup_request(ntp_message
*p
){
285 memset(p
, 0, sizeof(ntp_message
));
286 LI_SET(p
->flags
, LI_ALARM
);
288 MODE_SET(p
->flags
, MODE_CLIENT
);
290 p
->precision
=(int8_t)0xfa;
291 L16(p
->rtdelay
)=htons(1);
292 L16(p
->rtdisp
)=htons(1);
294 gettimeofday(&t
, NULL
);
295 TVtoNTP64(t
,p
->txts
);
298 /* select the "best" server from a list of servers, and return its index.
299 * this is done by filtering servers based on stratum, dispersion, and
300 * finally round-trip delay. */
301 int best_offset_server(const ntp_server_results
*slist
, int nservers
){
302 int i
=0, j
=0, cserver
=0, candidates
[5], csize
=0;
304 /* for each server */
305 for(cserver
=0; cserver
<nservers
; cserver
++){
306 /* sort out servers with error flags */
307 if ( LI(slist
[cserver
].flags
) != LI_NOWARNING
){
308 if (verbose
) printf("discarding peer id %d: flags=%d\n", cserver
, LI(slist
[cserver
].flags
));
312 /* compare it to each of the servers already in the candidate list */
313 for(i
=0; i
<csize
; i
++){
314 /* does it have an equal or better stratum? */
315 if(slist
[cserver
].stratum
<= slist
[i
].stratum
){
316 /* does it have an equal or better dispersion? */
317 if(slist
[cserver
].rtdisp
<= slist
[i
].rtdisp
){
318 /* does it have a better rtdelay? */
319 if(slist
[cserver
].rtdelay
< slist
[i
].rtdelay
){
326 /* if we haven't reached the current list's end, move everyone
327 * over one to the right, and insert the new candidate */
330 candidates
[j
]=candidates
[j
-1];
333 /* regardless, if they should be on the list... */
335 candidates
[i
]=cserver
;
337 /* otherwise discard the server */
339 DBG(printf("discarding peer id %d\n", cserver
));
344 DBG(printf("best server selected: peer %d\n", candidates
[0]));
345 return candidates
[0];
347 DBG(printf("no peers meeting synchronization criteria :(\n"));
352 /* do everything we need to get the total average offset
353 * - we use a certain amount of parallelization with poll() to ensure
354 * we don't waste time sitting around waiting for single packets.
355 * - we also "manually" handle resolving host names and connecting, because
356 * we have to do it in a way that our lazy macros don't handle currently :( */
357 double offset_request(const char *host
, int *status
){
358 int i
=0, j
=0, ga_result
=0, num_hosts
=0, *socklist
=NULL
, respnum
=0;
359 int servers_completed
=0, one_written
=0, one_read
=0, servers_readable
=0, best_index
=-1;
360 time_t now_time
=0, start_ts
=0;
361 ntp_message
*req
=NULL
;
362 double avg_offset
=0.;
363 struct timeval recv_time
;
364 struct addrinfo
*ai
=NULL
, *ai_tmp
=NULL
, hints
;
365 struct pollfd
*ufds
=NULL
;
366 ntp_server_results
*servers
=NULL
;
368 /* setup hints to only return results from getaddrinfo that we'd like */
369 memset(&hints
, 0, sizeof(struct addrinfo
));
370 hints
.ai_family
= address_family
;
371 hints
.ai_protocol
= IPPROTO_UDP
;
372 hints
.ai_socktype
= SOCK_DGRAM
;
374 /* fill in ai with the list of hosts resolved by the host name */
375 ga_result
= getaddrinfo(host
, "123", &hints
, &ai
);
377 die(STATE_UNKNOWN
, "error getting address for %s: %s\n",
378 host
, gai_strerror(ga_result
));
381 /* count the number of returned hosts, and allocate stuff accordingly */
382 for(ai_tmp
=ai
; ai_tmp
!=NULL
; ai_tmp
=ai_tmp
->ai_next
){ num_hosts
++; }
383 req
=(ntp_message
*)malloc(sizeof(ntp_message
)*num_hosts
);
384 if(req
==NULL
) die(STATE_UNKNOWN
, "can not allocate ntp message array");
385 socklist
=(int*)malloc(sizeof(int)*num_hosts
);
386 if(socklist
==NULL
) die(STATE_UNKNOWN
, "can not allocate socket array");
387 ufds
=(struct pollfd
*)malloc(sizeof(struct pollfd
)*num_hosts
);
388 if(ufds
==NULL
) die(STATE_UNKNOWN
, "can not allocate socket array");
389 servers
=(ntp_server_results
*)malloc(sizeof(ntp_server_results
)*num_hosts
);
390 if(servers
==NULL
) die(STATE_UNKNOWN
, "can not allocate server array");
391 memset(servers
, 0, sizeof(ntp_server_results
)*num_hosts
);
393 /* setup each socket for writing, and the corresponding struct pollfd */
396 socklist
[i
]=socket(ai_tmp
->ai_family
, SOCK_DGRAM
, IPPROTO_UDP
);
397 if(socklist
[i
] == -1) {
399 die(STATE_UNKNOWN
, "can not create new socket");
401 if(connect(socklist
[i
], ai_tmp
->ai_addr
, ai_tmp
->ai_addrlen
)){
402 die(STATE_UNKNOWN
, "can't create socket connection");
404 ufds
[i
].fd
=socklist
[i
];
405 ufds
[i
].events
=POLLIN
;
408 ai_tmp
= ai_tmp
->ai_next
;
411 /* now do AVG_NUM checks to each host. we stop before timeout/2 seconds
412 * have passed in order to ensure post-processing and jitter time. */
413 now_time
=start_ts
=time(NULL
);
414 while(servers_completed
<num_hosts
&& now_time
-start_ts
<= socket_timeout
/2){
415 /* loop through each server and find each one which hasn't
416 * been touched in the past second or so and is still lacking
417 * some responses. for each of these servers, send a new request,
418 * and update the "waiting" timestamp with the current time. */
422 for(i
=0; i
<num_hosts
; i
++){
423 if(servers
[i
].waiting
<now_time
&& servers
[i
].num_responses
<AVG_NUM
){
424 if(verbose
&& servers
[i
].waiting
!= 0) printf("re-");
425 if(verbose
) printf("sending request to peer %d\n", i
);
426 setup_request(&req
[i
]);
427 write(socklist
[i
], &req
[i
], sizeof(ntp_message
));
428 servers
[i
].waiting
=now_time
;
434 /* quickly poll for any sockets with pending data */
435 servers_readable
=poll(ufds
, num_hosts
, 100);
436 if(servers_readable
==-1){
437 perror("polling ntp sockets");
438 die(STATE_UNKNOWN
, "communication errors");
441 /* read from any sockets with pending data */
442 for(i
=0; servers_readable
&& i
<num_hosts
; i
++){
443 if(ufds
[i
].revents
&POLLIN
&& servers
[i
].num_responses
< AVG_NUM
){
445 printf("response from peer %d: ", i
);
448 read(ufds
[i
].fd
, &req
[i
], sizeof(ntp_message
));
449 gettimeofday(&recv_time
, NULL
);
450 DBG(print_ntp_message(&req
[i
]));
451 respnum
=servers
[i
].num_responses
++;
452 servers
[i
].offset
[respnum
]=calc_offset(&req
[i
], &recv_time
);
454 printf("offset %.10g\n", servers
[i
].offset
[respnum
]);
456 servers
[i
].stratum
=req
[i
].stratum
;
457 servers
[i
].rtdisp
=NTP32asDOUBLE(req
[i
].rtdisp
);
458 servers
[i
].rtdelay
=NTP32asDOUBLE(req
[i
].rtdelay
);
459 servers
[i
].waiting
=0;
460 servers
[i
].flags
=req
[i
].flags
;
463 if(servers
[i
].num_responses
==AVG_NUM
) servers_completed
++;
466 /* lather, rinse, repeat. */
470 die(STATE_CRITICAL
, "NTP CRITICAL: No response from NTP server\n");
473 /* now, pick the best server from the list */
474 best_index
=best_offset_server(servers
, num_hosts
);
476 *status
=STATE_CRITICAL
;
478 /* finally, calculate the average offset */
479 for(i
=0; i
<servers
[best_index
].num_responses
;i
++){
480 avg_offset
+=servers
[best_index
].offset
[j
];
482 avg_offset
/=servers
[best_index
].num_responses
;
486 /* FIXME: Not closing the socket to avoid re-use of the local port
487 * which can cause old NTP packets to be read instead of NTP control
488 * pactets in jitter_request(). THERE MUST BE ANOTHER WAY...
489 * for(j=0; j<num_hosts; j++){ close(socklist[j]); } */
496 if(verbose
) printf("overall average offset: %.10g\n", avg_offset
);
501 setup_control_request(ntp_control_message
*p
, uint8_t opcode
, uint16_t seq
){
502 memset(p
, 0, sizeof(ntp_control_message
));
503 LI_SET(p
->flags
, LI_NOWARNING
);
504 VN_SET(p
->flags
, VN_RESERVED
);
505 MODE_SET(p
->flags
, MODE_CONTROLMSG
);
506 OP_SET(p
->op
, opcode
);
508 /* Remaining fields are zero for requests */
511 /* XXX handle responses with the error bit set */
512 double jitter_request(const char *host
, int *status
){
513 int conn
=-1, i
, npeers
=0, num_candidates
=0, syncsource_found
=0;
514 int run
=0, min_peer_sel
=PEER_INCLUDED
, num_selected
=0, num_valid
=0;
515 int peers_size
=0, peer_offset
=0;
516 ntp_assoc_status_pair
*peers
=NULL
;
517 ntp_control_message req
;
518 const char *getvar
= "jitter";
519 double rval
= 0.0, jitter
= -1.0;
520 char *startofvalue
=NULL
, *nptr
=NULL
;
523 /* Long-winded explanation:
524 * Getting the jitter requires a number of steps:
525 * 1) Send a READSTAT request.
526 * 2) Interpret the READSTAT reply
527 * a) The data section contains a list of peer identifiers (16 bits)
528 * and associated status words (16 bits)
529 * b) We want the value of 0x06 in the SEL (peer selection) value,
530 * which means "current synchronizatin source". If that's missing,
531 * we take anything better than 0x04 (see the rfc for details) but
532 * set a minimum of warning.
533 * 3) Send a READVAR request for information on each peer identified
534 * in 2b greater than the minimum selection value.
535 * 4) Extract the jitter value from the data[] (it's ASCII)
537 my_udp_connect(server_address
, 123, &conn
);
539 /* keep sending requests until the server stops setting the
540 * REM_MORE bit, though usually this is only 1 packet. */
542 setup_control_request(&req
, OP_READSTAT
, 1);
543 DBG(printf("sending READSTAT request"));
544 write(conn
, &req
, SIZEOF_NTPCM(req
));
545 DBG(print_ntp_control_message(&req
));
546 /* Attempt to read the largest size packet possible */
547 req
.count
=htons(MAX_CM_SIZE
);
548 DBG(printf("recieving READSTAT response"))
549 read(conn
, &req
, SIZEOF_NTPCM(req
));
550 DBG(print_ntp_control_message(&req
));
551 /* Each peer identifier is 4 bytes in the data section, which
552 * we represent as a ntp_assoc_status_pair datatype.
554 peers_size
+=ntohs(req
.count
);
555 if((tmp
=realloc(peers
, peers_size
)) == NULL
)
556 free(peers
), die(STATE_UNKNOWN
, "can not (re)allocate 'peers' buffer\n");
558 memcpy((void*)((ptrdiff_t)peers
+peer_offset
), (void*)req
.data
, ntohs(req
.count
));
559 npeers
=peers_size
/sizeof(ntp_assoc_status_pair
);
560 peer_offset
+=ntohs(req
.count
);
561 } while(req
.op
&REM_MORE
);
563 /* first, let's find out if we have a sync source, or if there are
564 * at least some candidates. in the case of the latter we'll issue
565 * a warning but go ahead with the check on them. */
566 for (i
= 0; i
< npeers
; i
++){
567 if (PEER_SEL(peers
[i
].status
) >= PEER_INCLUDED
){
569 if(PEER_SEL(peers
[i
].status
) >= PEER_SYNCSOURCE
){
571 min_peer_sel
=PEER_SYNCSOURCE
;
575 if(verbose
) printf("%d candiate peers available\n", num_candidates
);
576 if(verbose
&& syncsource_found
) printf("synchronization source found\n");
577 if(! syncsource_found
){
578 *status
= STATE_WARNING
;
579 if(verbose
) printf("warning: no synchronization source found\n");
583 for (run
=0; run
<AVG_NUM
; run
++){
584 if(verbose
) printf("jitter run %d of %d\n", run
+1, AVG_NUM
);
585 for (i
= 0; i
< npeers
; i
++){
586 /* Only query this server if it is the current sync source */
587 if (PEER_SEL(peers
[i
].status
) >= min_peer_sel
){
589 setup_control_request(&req
, OP_READVAR
, 2);
590 req
.assoc
= peers
[i
].assoc
;
591 /* By spec, putting the variable name "jitter" in the request
592 * should cause the server to provide _only_ the jitter value.
593 * thus reducing net traffic, guaranteeing us only a single
594 * datagram in reply, and making intepretation much simpler
596 /* Older servers doesn't know what jitter is, so if we get an
597 * error on the first pass we redo it with "dispersion" */
598 strncpy(req
.data
, getvar
, MAX_CM_SIZE
-1);
599 req
.count
= htons(strlen(getvar
));
600 DBG(printf("sending READVAR request...\n"));
601 write(conn
, &req
, SIZEOF_NTPCM(req
));
602 DBG(print_ntp_control_message(&req
));
604 req
.count
= htons(MAX_CM_SIZE
);
605 DBG(printf("recieving READVAR response...\n"));
606 read(conn
, &req
, SIZEOF_NTPCM(req
));
607 DBG(print_ntp_control_message(&req
));
609 if(req
.op
&REM_ERROR
&& strstr(getvar
, "jitter")) {
610 if(verbose
) printf("The 'jitter' command failed (old ntp server?)\nRestarting with 'dispersion'...\n");
611 getvar
= "dispersion";
617 /* get to the float value */
619 printf("parsing jitter from peer %.2x: ", ntohs(peers
[i
].assoc
));
621 startofvalue
= strchr(req
.data
, '=');
622 if(startofvalue
!= NULL
) {
624 jitter
= strtod(startofvalue
, &nptr
);
626 if(startofvalue
== NULL
|| startofvalue
==nptr
){
627 printf("warning: unable to read server jitter response.\n");
628 *status
= STATE_WARNING
;
630 if(verbose
) printf("%g\n", jitter
);
637 printf("jitter parsed from %d/%d peers\n", num_valid
, num_selected
);
641 rval
= num_valid
? rval
/ num_valid
: -1.0;
644 if(peers
!=NULL
) free(peers
);
645 /* If we return -1.0, it means no synchronization source was found */
649 int process_arguments(int argc
, char **argv
){
652 static struct option longopts
[] = {
653 {"version", no_argument
, 0, 'V'},
654 {"help", no_argument
, 0, 'h'},
655 {"verbose", no_argument
, 0, 'v'},
656 {"use-ipv4", no_argument
, 0, '4'},
657 {"use-ipv6", no_argument
, 0, '6'},
658 {"warning", required_argument
, 0, 'w'},
659 {"critical", required_argument
, 0, 'c'},
660 {"jwarn", required_argument
, 0, 'j'},
661 {"jcrit", required_argument
, 0, 'k'},
662 {"timeout", required_argument
, 0, 't'},
663 {"hostname", required_argument
, 0, 'H'},
672 c
= getopt_long (argc
, argv
, "Vhv46w:c:j:k:t:H:", longopts
, &option
);
673 if (c
== -1 || c
== EOF
|| c
== 1)
682 print_revision(progname
, revision
);
689 owarn
= atof(optarg
);
692 ocrit
= atof(optarg
);
696 jwarn
= atof(optarg
);
700 jcrit
= atof(optarg
);
703 if(is_host(optarg
) == FALSE
)
704 usage2(_("Invalid hostname/address"), optarg
);
705 server_address
= strdup(optarg
);
708 socket_timeout
=atoi(optarg
);
711 address_family
= AF_INET
;
715 address_family
= AF_INET6
;
717 usage4 (_("IPv6 support not available"));
721 /* print short usage statement if args not parsable */
728 usage4(_("Critical offset should be larger than warning offset"));
732 usage4(_("Critical jitter should be larger than warning jitter"));
735 if(server_address
== NULL
){
736 usage4(_("Hostname was not supplied"));
742 char *perfd_offset (double offset
)
744 return fperfdata ("offset", offset
, "s",
750 char *perfd_jitter (double jitter
)
752 return fperfdata ("jitter", jitter
, "s",
758 int main(int argc
, char *argv
[]){
759 int result
, offset_result
, jitter_result
;
760 double offset
=0, jitter
=0;
761 char *result_line
, *perfdata_line
;
763 result
=offset_result
=jitter_result
=STATE_UNKNOWN
;
765 if (process_arguments (argc
, argv
) == ERROR
)
766 usage4 (_("Could not parse arguments"));
768 /* initialize alarm signal handling */
769 signal (SIGALRM
, socket_timeout_alarm_handler
);
771 /* set socket timeout */
772 alarm (socket_timeout
);
774 offset
= offset_request(server_address
, &offset_result
);
775 if(fabs(offset
) > ocrit
){
776 result
= STATE_CRITICAL
;
777 } else if(fabs(offset
) > owarn
) {
778 result
= STATE_WARNING
;
782 result
=max_state(result
, offset_result
);
784 /* If not told to check the jitter, we don't even send packets.
785 * jitter is checked using NTP control packets, which not all
786 * servers recognize. Trying to check the jitter on OpenNTPD
787 * (for example) will result in an error
790 jitter
=jitter_request(server_address
, &jitter_result
);
792 result
= max_state(result
, STATE_CRITICAL
);
793 } else if(jitter
> jwarn
) {
794 result
= max_state(result
, STATE_WARNING
);
795 } else if(jitter
== -1.0 && result
== STATE_OK
){
796 /* -1 indicates that we couldn't calculate the jitter
797 * Only overrides STATE_OK from the offset */
798 result
= STATE_UNKNOWN
;
801 result
=max_state(result
, jitter_result
);
804 case STATE_CRITICAL
:
805 asprintf(&result_line
, "NTP CRITICAL:");
808 asprintf(&result_line
, "NTP WARNING:");
811 asprintf(&result_line
, "NTP OK:");
814 asprintf(&result_line
, "NTP UNKNOWN:");
817 if(offset_result
==STATE_CRITICAL
){
818 asprintf(&result_line
, "%s %s", result_line
, _("Offset unknown"));
819 asprintf(&perfdata_line
, "");
821 if(offset_result
==STATE_WARNING
){
822 asprintf(&result_line
, "%s %s", result_line
, _("Unable to fully sample sync server"));
824 asprintf(&result_line
, "%s Offset %.10g secs", result_line
, offset
);
825 asprintf(&perfdata_line
, "%s", perfd_offset(offset
));
828 asprintf(&result_line
, "%s, jitter=%f", result_line
, jitter
);
829 asprintf(&perfdata_line
, "%s %s", perfdata_line
, perfd_jitter(jitter
));
831 printf("%s|%s\n", result_line
, perfdata_line
);
833 if(server_address
!=NULL
) free(server_address
);
839 void print_help(void){
840 print_revision(progname
, revision
);
842 printf ("Copyright (c) 2006 Sean Finney\n");
843 printf (COPYRIGHT
, copyright
, email
);
845 printf ("%s\n", _("This plugin checks the selected ntp server"));
850 printf (_(UT_HELP_VRSN
));
851 printf (_(UT_HOST_PORT
), 'p', "123");
852 printf (" %s\n", "-w, --warning=DOUBLE");
853 printf (" %s\n", _("Offset to result in warning status (seconds)"));
854 printf (" %s\n", "-c, --critical=DOUBLE");
855 printf (" %s\n", _("Offset to result in critical status (seconds)"));
856 printf (" %s\n", "-j, --warning=DOUBLE");
857 printf (" %s\n", _("Warning value for jitter"));
858 printf (" %s\n", "-k, --critical=DOUBLE");
859 printf (" %s\n", _("Critical value for jitter"));
860 printf (_(UT_TIMEOUT
), DEFAULT_SOCKET_TIMEOUT
);
861 printf (_(UT_VERBOSE
));
862 printf (_(UT_SUPPORT
));
868 printf (_("Usage:"));
869 printf("%s -H <host> [-w <warn>] [-c <crit>] [-j <warn>] [-k <crit>] [-v verbose]\n", progname
);