From 22f47fc0c51266bd25491da30fd165a845a1391b Mon Sep 17 00:00:00 2001 From: Sven Nierlein Date: Mon, 21 Jan 2019 17:41:02 +0100 Subject: tests: make check_snmp test more reliable The check_snmp rate tests depend on the exact amount of time spend between the plugin runs and will fail on busy machines, ex. the ci servers. Using faketime mitigates this issue and also removes all the sleeps. Signed-off-by: Sven Nierlein --- plugins/tests/check_snmp.t | 110 ++++++++++++++++++++++++--------------------- 1 file changed, 58 insertions(+), 52 deletions(-) (limited to 'plugins/tests/check_snmp.t') diff --git a/plugins/tests/check_snmp.t b/plugins/tests/check_snmp.t index 73a68b20..85d6bf55 100755 --- a/plugins/tests/check_snmp.t +++ b/plugins/tests/check_snmp.t @@ -7,6 +7,7 @@ use strict; use Test::More; use NPTest; use FindBin qw($Bin); +use POSIX qw/strftime/; my $tests = 67; # Check that all dependent modules are available @@ -37,6 +38,7 @@ if ($@) { my $port_snmp = 16100 + int(rand(100)); +my $faketime = -x '/usr/bin/faketime' ? 1 : 0; # Start up server my @pids; @@ -118,77 +120,81 @@ like($res->output, '/'.quotemeta('SNMP OK - And now have fun with with this: \"C "And now have fun with with this: \"C:\\\\\" because we\'re not done yet!"').'/m', "Attempt to confuse parser No.3"); -system("rm -f ".$ENV{'MP_STATE_PATH'}."/check_snmp/*"); -$res = NPTest->testCmd( "./check_snmp -H 127.0.0.1 -C public -p $port_snmp -o .1.3.6.1.4.1.8072.3.2.67.10 --rate -w 600" ); -is($res->return_code, 0, "Returns OK"); -is($res->output, "No previous data to calculate rate - assume okay"); +system("rm -f ".$ENV{'MP_STATE_PATH'}."/*/check_snmp/*"); -# Need to sleep, otherwise duration=0 -sleep 1; +# run rate checks with faketime. rate checks depend on the exact amount of time spend between the +# plugin runs which may fail on busy machines. +# using faketime removes this race condition and also saves all the sleeps in between. +SKIP: { + skip "No faketime binary found", 28 if !$faketime; -$res = NPTest->testCmd( "./check_snmp -H 127.0.0.1 -C public -p $port_snmp -o .1.3.6.1.4.1.8072.3.2.67.10 --rate -w 600" ); -is($res->return_code, 1, "WARNING - due to going above rate calculation" ); -is($res->output, "SNMP RATE WARNING - *666* | iso.3.6.1.4.1.8072.3.2.67.10=666;600 "); + my $ts = time(); + $res = NPTest->testCmd("LC_TIME=C TZ=UTC faketime -f '".strftime("%Y-%m-%d %H:%M:%S", localtime($ts))."' ./check_snmp -H 127.0.0.1 -C public -p $port_snmp -o .1.3.6.1.4.1.8072.3.2.67.10 --rate -w 600" ); + is($res->return_code, 0, "Returns OK"); + is($res->output, "No previous data to calculate rate - assume okay"); -$res = NPTest->testCmd( "./check_snmp -H 127.0.0.1 -C public -p $port_snmp -o .1.3.6.1.4.1.8072.3.2.67.10 --rate -w 600" ); -is($res->return_code, 3, "UNKNOWN - basically the divide by zero error" ); -is($res->output, "Time duration between plugin calls is invalid"); + # test rate 1 second later + $res = NPTest->testCmd("LC_TIME=C TZ=UTC faketime -f '".strftime("%Y-%m-%d %H:%M:%S", localtime($ts+1))."' ./check_snmp -H 127.0.0.1 -C public -p $port_snmp -o .1.3.6.1.4.1.8072.3.2.67.10 --rate -w 600" ); + is($res->return_code, 1, "WARNING - due to going above rate calculation" ); + is($res->output, "SNMP RATE WARNING - *666* | iso.3.6.1.4.1.8072.3.2.67.10=666;600 "); + # test rate with same time + $res = NPTest->testCmd("LC_TIME=C TZ=UTC faketime -f '".strftime("%Y-%m-%d %H:%M:%S", localtime($ts+1))."' ./check_snmp -H 127.0.0.1 -C public -p $port_snmp -o .1.3.6.1.4.1.8072.3.2.67.10 --rate -w 600" ); + is($res->return_code, 3, "UNKNOWN - basically the divide by zero error" ); + is($res->output, "Time duration between plugin calls is invalid"); -$res = NPTest->testCmd( "./check_snmp -H 127.0.0.1 -C public -p $port_snmp -o .1.3.6.1.4.1.8072.3.2.67.10 --rate -l inoctets" ); -is($res->return_code, 0, "OK for first call" ); -is($res->output, "No previous data to calculate rate - assume okay" ); -# Need to sleep, otherwise duration=0 -sleep 1; + $res = NPTest->testCmd("LC_TIME=C TZ=UTC faketime -f '".strftime("%Y-%m-%d %H:%M:%S", localtime($ts))."' ./check_snmp -H 127.0.0.1 -C public -p $port_snmp -o .1.3.6.1.4.1.8072.3.2.67.10 --rate -l inoctets" ); + is($res->return_code, 0, "OK for first call" ); + is($res->output, "No previous data to calculate rate - assume okay" ); -$res = NPTest->testCmd( "./check_snmp -H 127.0.0.1 -C public -p $port_snmp -o .1.3.6.1.4.1.8072.3.2.67.10 --rate -l inoctets" ); -is($res->return_code, 0, "OK as no thresholds" ); -is($res->output, "SNMP RATE OK - inoctets 666 | inoctets=666 ", "Check label"); + # test rate 1 second later + $res = NPTest->testCmd("LC_TIME=C TZ=UTC faketime -f '".strftime("%Y-%m-%d %H:%M:%S", localtime($ts+1))."' ./check_snmp -H 127.0.0.1 -C public -p $port_snmp -o .1.3.6.1.4.1.8072.3.2.67.10 --rate -l inoctets" ); + is($res->return_code, 0, "OK as no thresholds" ); + is($res->output, "SNMP RATE OK - inoctets 666 | inoctets=666 ", "Check label"); -sleep 2; + # test rate 3 seconds later + $res = NPTest->testCmd("LC_TIME=C TZ=UTC faketime -f '".strftime("%Y-%m-%d %H:%M:%S", localtime($ts+3))."' ./check_snmp -H 127.0.0.1 -C public -p $port_snmp -o .1.3.6.1.4.1.8072.3.2.67.10 --rate -l inoctets" ); + is($res->return_code, 0, "OK as no thresholds" ); + is($res->output, "SNMP RATE OK - inoctets 333 | inoctets=333 ", "Check rate decreases due to longer interval"); -$res = NPTest->testCmd( "./check_snmp -H 127.0.0.1 -C public -p $port_snmp -o .1.3.6.1.4.1.8072.3.2.67.10 --rate -l inoctets" ); -is($res->return_code, 0, "OK as no thresholds" ); -is($res->output, "SNMP RATE OK - inoctets 333 | inoctets=333 ", "Check rate decreases due to longer interval"); + # label performance data check + $res = NPTest->testCmd( "./check_snmp -H 127.0.0.1 -C public -p $port_snmp -o .1.3.6.1.4.1.8072.3.2.67.10 -l test" ); + is($res->return_code, 0, "OK as no thresholds" ); + is($res->output, "SNMP OK - test 67996 | test=67996c ", "Check label"); -# label performance data check -$res = NPTest->testCmd( "./check_snmp -H 127.0.0.1 -C public -p $port_snmp -o .1.3.6.1.4.1.8072.3.2.67.10 -l test" ); -is($res->return_code, 0, "OK as no thresholds" ); -is($res->output, "SNMP OK - test 67996 | test=67996c ", "Check label"); + $res = NPTest->testCmd( "./check_snmp -H 127.0.0.1 -C public -p $port_snmp -o .1.3.6.1.4.1.8072.3.2.67.10 -l \"test'test\"" ); + is($res->return_code, 0, "OK as no thresholds" ); + is($res->output, "SNMP OK - test'test 68662 | \"test'test\"=68662c ", "Check label"); -$res = NPTest->testCmd( "./check_snmp -H 127.0.0.1 -C public -p $port_snmp -o .1.3.6.1.4.1.8072.3.2.67.10 -l \"test'test\"" ); -is($res->return_code, 0, "OK as no thresholds" ); -is($res->output, "SNMP OK - test'test 68662 | \"test'test\"=68662c ", "Check label"); + $res = NPTest->testCmd( "./check_snmp -H 127.0.0.1 -C public -p $port_snmp -o .1.3.6.1.4.1.8072.3.2.67.10 -l 'test\"test'" ); + is($res->return_code, 0, "OK as no thresholds" ); + is($res->output, "SNMP OK - test\"test 69328 | 'test\"test'=69328c ", "Check label"); -$res = NPTest->testCmd( "./check_snmp -H 127.0.0.1 -C public -p $port_snmp -o .1.3.6.1.4.1.8072.3.2.67.10 -l 'test\"test'" ); -is($res->return_code, 0, "OK as no thresholds" ); -is($res->output, "SNMP OK - test\"test 69328 | 'test\"test'=69328c ", "Check label"); + $res = NPTest->testCmd( "./check_snmp -H 127.0.0.1 -C public -p $port_snmp -o .1.3.6.1.4.1.8072.3.2.67.10 -l test -O" ); + is($res->return_code, 0, "OK as no thresholds" ); + is($res->output, "SNMP OK - test 69994 | iso.3.6.1.4.1.8072.3.2.67.10=69994c ", "Check label"); -$res = NPTest->testCmd( "./check_snmp -H 127.0.0.1 -C public -p $port_snmp -o .1.3.6.1.4.1.8072.3.2.67.10 -l test -O" ); -is($res->return_code, 0, "OK as no thresholds" ); -is($res->output, "SNMP OK - test 69994 | iso.3.6.1.4.1.8072.3.2.67.10=69994c ", "Check label"); + $res = NPTest->testCmd( "./check_snmp -H 127.0.0.1 -C public -p $port_snmp -o .1.3.6.1.4.1.8072.3.2.67.10" ); + is($res->return_code, 0, "OK as no thresholds" ); + is($res->output, "SNMP OK - 70660 | iso.3.6.1.4.1.8072.3.2.67.10=70660c ", "Check label"); -$res = NPTest->testCmd( "./check_snmp -H 127.0.0.1 -C public -p $port_snmp -o .1.3.6.1.4.1.8072.3.2.67.10" ); -is($res->return_code, 0, "OK as no thresholds" ); -is($res->output, "SNMP OK - 70660 | iso.3.6.1.4.1.8072.3.2.67.10=70660c ", "Check label"); + $res = NPTest->testCmd( "./check_snmp -H 127.0.0.1 -C public -p $port_snmp -o .1.3.6.1.4.1.8072.3.2.67.10 -l 'test test'" ); + is($res->return_code, 0, "OK as no thresholds" ); + is($res->output, "SNMP OK - test test 71326 | 'test test'=71326c ", "Check label"); -$res = NPTest->testCmd( "./check_snmp -H 127.0.0.1 -C public -p $port_snmp -o .1.3.6.1.4.1.8072.3.2.67.10 -l 'test test'" ); -is($res->return_code, 0, "OK as no thresholds" ); -is($res->output, "SNMP OK - test test 71326 | 'test test'=71326c ", "Check label"); + $res = NPTest->testCmd("LC_TIME=C TZ=UTC faketime -f '".strftime("%Y-%m-%d %H:%M:%S", localtime($ts))."' ./check_snmp -H 127.0.0.1 -C public -p $port_snmp -o .1.3.6.1.4.1.8072.3.2.67.10 --rate -l inoctets_per_minute --rate-multiplier=60" ); + is($res->return_code, 0, "OK for first call" ); + is($res->output, "No previous data to calculate rate - assume okay" ); -$res = NPTest->testCmd( "./check_snmp -H 127.0.0.1 -C public -p $port_snmp -o .1.3.6.1.4.1.8072.3.2.67.10 --rate -l inoctets_per_minute --rate-multiplier=60" ); -is($res->return_code, 0, "OK for first call" ); -is($res->output, "No previous data to calculate rate - assume okay" ); - -# Need to sleep, otherwise duration=0 -sleep 1; + # test 1 second later + $res = NPTest->testCmd("LC_TIME=C TZ=UTC faketime -f '".strftime("%Y-%m-%d %H:%M:%S", localtime($ts+1))."' ./check_snmp -H 127.0.0.1 -C public -p $port_snmp -o .1.3.6.1.4.1.8072.3.2.67.10 --rate -l inoctets_per_minute --rate-multiplier=60" ); + is($res->return_code, 0, "OK as no thresholds" ); + is($res->output, "SNMP RATE OK - inoctets_per_minute 39960 | inoctets_per_minute=39960 ", "Checking multiplier"); +}; -$res = NPTest->testCmd( "./check_snmp -H 127.0.0.1 -C public -p $port_snmp -o .1.3.6.1.4.1.8072.3.2.67.10 --rate -l inoctets_per_minute --rate-multiplier=60" ); -is($res->return_code, 0, "OK as no thresholds" ); -is($res->output, "SNMP RATE OK - inoctets_per_minute 39960 | inoctets_per_minute=39960 ", "Checking multiplier"); $res = NPTest->testCmd( "./check_snmp -H 127.0.0.1 -C public -p $port_snmp -o .1.3.6.1.4.1.8072.3.2.67.11 -s '\"stringtests\"'" ); -- cgit v1.2.3-74-g34f1 From 024d268386353133af1a9ff5c0b5879397c19b1c Mon Sep 17 00:00:00 2001 From: Sven Nierlein Date: Wed, 17 Nov 2021 11:58:41 +0100 Subject: check_snmp: fix performance thresholds when using multiple oids when using check_snmp with multiple oids it simply printed the unparsed content from -w/-c into the thresholds for each oid. So each oid contained the hole -w from all oids. ./check_snmp ... -o iso.3.6.1.2.1.25.1.3.0,iso.3.6.1.2.1.25.1.5.0 -w '1,2' -c '3,4' before: SNMP ... | HOST-RESOURCES-MIB::hrSystemInitialLoadDevice.0=393216;1,2;3,4 HOST-RESOURCES-MIB::hrSystemNumUsers.0=24;1,2;3,4 after: SNMP ... | HOST-RESOURCES-MIB::hrSystemInitialLoadDevice.0=393216;1;3 HOST-RESOURCES-MIB::hrSystemNumUsers.0=24;2;4 This also applies to fixed thresholds since check_snmp translates negative infinities from: '~:-1' to '@-1:~' --- lib/utils_base.c | 2 +- lib/utils_base.h | 1 + plugins/check_snmp.c | 9 +++++++-- plugins/tests/check_snmp.t | 17 ++++++++++++++--- 4 files changed, 23 insertions(+), 6 deletions(-) (limited to 'plugins/tests/check_snmp.t') diff --git a/lib/utils_base.c b/lib/utils_base.c index fd7058da..08fa215c 100644 --- a/lib/utils_base.c +++ b/lib/utils_base.c @@ -128,6 +128,7 @@ range temp_range->end = 0; temp_range->end_infinity = TRUE; temp_range->alert_on = OUTSIDE; + temp_range->text = strdup(str); if (str[0] == '@') { temp_range->alert_on = INSIDE; @@ -706,4 +707,3 @@ void np_state_write_string(time_t data_time, char *data_string) { np_free(temp_file); } - diff --git a/lib/utils_base.h b/lib/utils_base.h index d7e7dffa..9482f23b 100644 --- a/lib/utils_base.h +++ b/lib/utils_base.h @@ -23,6 +23,7 @@ typedef struct range_struct { double end; int end_infinity; int alert_on; /* OUTSIDE (default) or INSIDE */ + char* text; /* original unparsed text input */ } range; typedef struct thresholds_struct { diff --git a/plugins/check_snmp.c b/plugins/check_snmp.c index afc568b2..58d46b11 100644 --- a/plugins/check_snmp.c +++ b/plugins/check_snmp.c @@ -468,6 +468,9 @@ main (int argc, char **argv) /* Process this block for numeric comparisons */ /* Make some special values,like Timeticks numeric only if a threshold is defined */ if (thlds[i]->warning || thlds[i]->critical || calculate_rate) { + if (verbose > 2) { + print_thresholds(" thresholds", thlds[i]); + } ptr = strpbrk (show, "-0123456789"); if (ptr == NULL) die (STATE_UNKNOWN,_("No valid data returned (%s)\n"), show); @@ -581,14 +584,16 @@ main (int argc, char **argv) if (warning_thresholds) { strncat(perfstr, ";", sizeof(perfstr)-strlen(perfstr)-1); - strncat(perfstr, warning_thresholds, sizeof(perfstr)-strlen(perfstr)-1); + if(thlds[i]->warning && thlds[i]->warning->text) + strncat(perfstr, thlds[i]->warning->text, sizeof(perfstr)-strlen(perfstr)-1); } if (critical_thresholds) { if (!warning_thresholds) strncat(perfstr, ";", sizeof(perfstr)-strlen(perfstr)-1); strncat(perfstr, ";", sizeof(perfstr)-strlen(perfstr)-1); - strncat(perfstr, critical_thresholds, sizeof(perfstr)-strlen(perfstr)-1); + if(thlds[i]->critical && thlds[i]->critical->text) + strncat(perfstr, thlds[i]->critical->text, sizeof(perfstr)-strlen(perfstr)-1); } strncat(perfstr, " ", sizeof(perfstr)-strlen(perfstr)-1); diff --git a/plugins/tests/check_snmp.t b/plugins/tests/check_snmp.t index 85d6bf55..0a77fa8a 100755 --- a/plugins/tests/check_snmp.t +++ b/plugins/tests/check_snmp.t @@ -9,7 +9,7 @@ use NPTest; use FindBin qw($Bin); use POSIX qw/strftime/; -my $tests = 67; +my $tests = 73; # Check that all dependent modules are available eval { require NetSNMP::OID; @@ -251,9 +251,20 @@ is($res->output, 'SNMP CRITICAL - *-4* | iso.3.6.1.4.1.8072.3.2.67.17=-4;-2:;-3: $res = NPTest->testCmd( "./check_snmp -H 127.0.0.1 -C public -p $port_snmp -o .1.3.6.1.4.1.8072.3.2.67.18 -c '~:-6.5'" ); is($res->return_code, 0, "Negative float OK" ); -is($res->output, 'SNMP OK - -6.6 | iso.3.6.1.4.1.8072.3.2.67.18=-6.6;;~:-6.5 ', "Negative float OK output" ); +is($res->output, 'SNMP OK - -6.6 | iso.3.6.1.4.1.8072.3.2.67.18=-6.6;;@-6.5:~ ', "Negative float OK output" ); $res = NPTest->testCmd( "./check_snmp -H 127.0.0.1 -C public -p $port_snmp -o .1.3.6.1.4.1.8072.3.2.67.18 -w '~:-6.65' -c '~:-6.55'" ); is($res->return_code, 1, "Negative float WARNING" ); -is($res->output, 'SNMP WARNING - *-6.6* | iso.3.6.1.4.1.8072.3.2.67.18=-6.6;~:-6.65;~:-6.55 ', "Negative float WARNING output" ); +is($res->output, 'SNMP WARNING - *-6.6* | iso.3.6.1.4.1.8072.3.2.67.18=-6.6;@-6.65:~;@-6.55:~ ', "Negative float WARNING output" ); +$res = NPTest->testCmd( "./check_snmp -H 127.0.0.1 -C public -p $port_snmp -o .1.3.6.1.4.1.8072.3.2.67.10,.1.3.6.1.4.1.8072.3.2.67.17 -w '1:100000,-10:20' -c '2:200000,-20:30'" ); +is($res->return_code, 0, "Multiple OIDs with thresholds" ); +like($res->output, '/SNMP OK - \d+ -4 | iso.3.6.1.4.1.8072.3.2.67.10=\d+c;1:100000;2:200000 iso.3.6.1.4.1.8072.3.2.67.17=-4;-10:20;-20:30/', "Multiple OIDs with thresholds output" ); + +$res = NPTest->testCmd( "./check_snmp -H 127.0.0.1 -C public -p $port_snmp -o .1.3.6.1.4.1.8072.3.2.67.10,.1.3.6.1.4.1.8072.3.2.67.17 -w '1:100000,-1:2' -c '2:200000,-20:30'" ); +is($res->return_code, 1, "Multiple OIDs with thresholds" ); +like($res->output, '/SNMP WARNING - \d+ \*-4\* | iso.3.6.1.4.1.8072.3.2.67.10=\d+c;1:100000;2:200000 iso.3.6.1.4.1.8072.3.2.67.17=-4;-10:20;-20:30/', "Multiple OIDs with thresholds output" ); + +$res = NPTest->testCmd( "./check_snmp -H 127.0.0.1 -C public -p $port_snmp -o .1.3.6.1.4.1.8072.3.2.67.10,.1.3.6.1.4.1.8072.3.2.67.17 -w 1,2 -c 1" ); +is($res->return_code, 2, "Multiple OIDs with some thresholds" ); +like($res->output, '/SNMP CRITICAL - \*\d+\* \*-4\* | iso.3.6.1.4.1.8072.3.2.67.10=\d+c;1;2 iso.3.6.1.4.1.8072.3.2.67.17=-4;;/', "Multiple OIDs with thresholds output" ); -- cgit v1.2.3-74-g34f1