Hi,
I was wondering some days ago, how to remove some metrics, since I'd like only
to see some values on the webpage.
I think I've found a solution in changing the code. If there is any possibility
in configuring and not in patching, please drop me a note!
I went to the source directory 2.5.6 into the directory gmond and modified the
files key_metrics.h and metric.h.
I removed several lines with the definietions of the metrics. (The
corresponding lines in both files)
I attatched the files for your info.
Regards
Cornelius
#ifndef KEY_METRICS_H
#define KEY_METRICS_H 1
enum {
user_defined,
/*
* These are my configuration metrics which don't change between reboots
*/
cpu_num,
cpu_speed,
mem_total,
swap_total,
boottime,
sys_clock,
machine_type,
os_name,
os_release,
/*
* These are my state metrics which are always a changin' changin'
*/
cpu_user,
cpu_nice,
cpu_system,
cpu_idle,
/* internal.. ignore */
gexec,
heartbeat,
mtu,
location,
/*
* added by swagner for extended solaris monitoring.
*/
#ifdef SOLARIS
cpu_wio,
bread_sec,
bwrite_sec,
lread_sec,
lwrite_sec,
phread_sec,
phwrite_sec,
rcache,
wcache,
#endif
#ifdef LINUX
#endif
#ifdef HPUX
cpu_intr,
cpu_ssys,
cpu_wait,
cpu_arm,
cpu_rm,
cpu_avm,
cpu_vm,
#endif
#ifdef FREEBSD
bytes_in,
bytes_out,
pkts_in,
pkts_out,
disk_total,
disk_free,
part_max_used,
#endif
num_key_metrics
} key_metrics;
#endif
#ifndef METRIC_H
#define METRIC_H 1
#include <ganglia.h>
#include "metric_typedefs.h"
#include "key_metrics.h"
/* SEE key_metrics.h for list of key metrics */
/* If you add a value to the key_metrics enum then you need to add
* its func here as well. They need to be of type "metric_func" defined
* above. These functions are all added to the metric_<os>.c file along
* with the metric_init() function. The metric_init() function is only
* called once.
*/
extern g_val_t metric_init(void);
extern g_val_t cpu_num_func(void);
extern g_val_t cpu_speed_func(void);
extern g_val_t mem_total_func(void);
extern g_val_t swap_total_func(void);
extern g_val_t boottime_func(void);
extern g_val_t sys_clock_func(void);
extern g_val_t machine_type_func(void);
extern g_val_t os_name_func(void);
extern g_val_t os_release_func(void);
extern g_val_t mtu_func(void);
extern g_val_t cpu_user_func(void);
extern g_val_t cpu_nice_func(void);
extern g_val_t cpu_system_func(void);
extern g_val_t cpu_idle_func(void);
extern g_val_t cpu_aidle_func(void);
extern g_val_t load_one_func(void);
extern g_val_t load_five_func(void);
extern g_val_t load_fifteen_func(void);
extern g_val_t proc_run_func(void);
extern g_val_t proc_total_func(void);
extern g_val_t mem_free_func(void);
extern g_val_t mem_shared_func(void);
extern g_val_t mem_buffers_func(void);
extern g_val_t mem_cached_func(void);
extern g_val_t swap_free_func(void);
extern g_val_t gexec_func(void);
extern g_val_t heartbeat_func(void);
extern g_val_t mtu_func(void);
extern g_val_t location_func(void);
/* the following are additional internal metrics added by swagner
* what for the monitoring of buffer/linear read/writes on Solaris boxen.
* these are only valid on the solaris version of gmond v2.3.1b1,
* all others are untested. caveat haxor. :P
*/
#ifdef SOLARIS
extern g_val_t bread_sec_func(void);
extern g_val_t bwrite_sec_func(void);
extern g_val_t lread_sec_func(void);
extern g_val_t lwrite_sec_func(void);
extern g_val_t phread_sec_func(void);
extern g_val_t phwrite_sec_func(void);
extern g_val_t rcache_func(void);
extern g_val_t wcache_func(void);
extern g_val_t cpu_wio_func(void);
#endif
#ifdef LINUX
extern g_val_t bytes_in_func(void);
extern g_val_t bytes_out_func(void);
extern g_val_t pkts_in_func(void);
extern g_val_t pkts_out_func(void);
extern g_val_t disk_total_func(void);
extern g_val_t disk_free_func(void);
extern g_val_t part_max_used_func(void);
#endif
#ifdef HPUX
extern g_val_t cpu_wait_func(void);
extern g_val_t cpu_intr_func(void);
extern g_val_t cpu_ssys_func(void);
extern g_val_t mem_rm_func(void);
extern g_val_t mem_arm_func(void);
extern g_val_t mem_vm_func(void);
extern g_val_t mem_avm_func(void);
#endif
#ifdef FREEBSD
extern g_val_t bytes_in_func(void);
extern g_val_t bytes_out_func(void);
extern g_val_t pkts_in_func(void);
extern g_val_t pkts_out_func(void);
extern g_val_t disk_total_func(void);
extern g_val_t disk_free_func(void);
extern g_val_t part_max_used_func(void);
#endif
#define INIT 0, 0, {0}, {0}
#define KEY(NAME) { #NAME, NAME ##_func, INIT
/*
* All behavior of the gmond is tweak'd by this array..
*
* Each metric is checked during some random interval between "check_min" and
* "check_max" (assuming the threshold value is not negative). If the metric is
* "threshold" greater than the previous metric value then the key_metric
* (listed on the left) is multicast. If there has been no multicast
* during some random interval from "mcast_min" to "mcast_max" then a multicast
* is sent regardless of the metric value.
*
* NOTE: it is critical that the key matches the key in the enum key_metrics
above
*/
metric_t metric[num_key_metrics] = {
/* threshold check_max mcast_max units */
/* key check_min mcast_min type fmt */
/*------------------------------------------------------------------*/
/* user_defined does nothing */
{"user_defined",metric_init,
INIT, -1, -1, -1, -1, -1, g_int32, "", ""},
/* CONFIGURATION METRICS */
KEY(cpu_num), -1, -1, -1, 900, 1200, g_uint16, "","%hu"},
KEY(cpu_speed), -1, -1, -1, 900, 1200, g_uint32, "MHz", "%hu"},
KEY(mem_total), -1, -1, -1, 900, 1200, g_uint32, "KB", "%u" },
KEY(swap_total), -1, -1, -1, 900, 1200, g_uint32, "KB", "%u" },
KEY(boottime), -1, -1, -1, 900, 1200, g_timestamp, "s", "%u" },
KEY(sys_clock), -1, -1, -1, 900, 1200, g_timestamp, "s", "%u" },
KEY(machine_type), -1, -1, -1, 900, 1200, g_string, "", "%s" },
KEY(os_name), -1, -1, -1, 900, 1200, g_string, "", "%s" },
KEY(os_release), -1, -1, -1, 900, 1200, g_string, "", "%s" },
/* STATE METRICS */
KEY(cpu_user), 1, 15, 20, 60, 90, g_float, "%", "%.1f"},
KEY(cpu_nice), 1, 15, 20, 60, 90, g_float, "%", "%.1f"},
KEY(cpu_system), 1, 15, 20, 60, 90, g_float, "%", "%.1f"},
KEY(cpu_idle), 5, 15, 20, 60, 90, g_float, "%", "%.1f"},
/* gmond internals */
KEY(gexec), -1, -1, -1, 180, 300, g_string, "", "%s"},
KEY(heartbeat), -1, -1, -1, 10, 20, g_uint32, "", "%u"},
KEY(mtu), -1, -1, -1, 900, 1200, g_uint32, "B", "%u" },
KEY(location), -1, -1, -1, 900, 1200, g_string, "(x,y,z)", "%s" }
#ifdef SOLARIS
,
KEY(cpu_wio), 5, 850, 950, 3400, 3800, g_float, "%", "%.1f"},
/* buffer reads and writes, adjusted per second */
KEY(bread_sec), 1, 15, 20, 60, 90, g_float, "", "%.2f" },
KEY(bwrite_sec), 1, 15, 20, 60, 90, g_float, "", "%.2f" },
/* linear reads/writes, adjusted per second */
KEY(lread_sec),1, 15, 20, 60, 90, g_float, "", "%.2f" },
KEY(lwrite_sec), 1, 15, 20, 60, 90, g_float, "", "%.2f" },
/* cache hit ratios, rcache = 1 - (bread/lread), wcache = 1 - (bwrite/lwrite) */
KEY(rcache), 1, 15, 20, 60, 90, g_float, "%", "%.1f" },
KEY(wcache), 1, 15, 20, 60, 90, g_float, "%", "%.1f" },
/* physical reads/writes per second. */
KEY(phread_sec), 1, 15, 20, 60, 90, g_float, "", "%.2f" },
KEY(phwrite_sec), 1, 15, 20, 60, 90, g_float, "", "%.2f" }
#endif
#ifdef LINUX
,
/* (sacerdoti) Experiments have shown gmon resting bandwidth is around 2KB/s for
* a 128-node cluster. We set the value thresh of these metrics to double that.
*/
/*
* The amount of disk space could change - hot-swap, mounts, etc. check:
30-60min. */
KEY(disk_total), 1, 1800, 3600, 900, 1200, g_double, "GB", "%.3f" },
KEY(disk_free), 1, 30, 40, 120, 180, g_double, "GB", "%.3f" },
KEY(part_max_used), 1, 30, 40, 120, 180, g_float, "%", "%.1f" }
#endif
#ifdef HPUX
,
KEY(cpu_intr), 1, 15, 20, 60, 90, g_float, "%", "%.1f"},
KEY(cpu_ssys), 1, 15, 20, 60, 90, g_float, "%", "%.1f"},
KEY(cpu_wait), 1, 15, 20, 60, 90, g_float, "%", "%.1f"},
KEY(mem_arm), 1024, 30, 40, 120, 180, g_uint32, "KB", "%u" },
KEY(mem_rm), 1024, 30, 40, 120, 180, g_uint32, "KB", "%u" },
KEY(mem_avm), 1024, 30, 40, 120, 180, g_uint32, "KB", "%u" },
KEY(mem_vm), 1024, 30, 40, 120, 180, g_uint32, "KB", "%u" }
#endif
#ifdef FREEBSD
,
/*
* (sacerdoti) Experiments have shown gmon resting bandwidth is around
* 2KB/s for a 128-node cluster. We set the value thresh of these
* metrics to double that.
*
*/
KEY(bytes_out), 4096, 30, 40, 200, 300, g_float, "bytes/sec", "%.2f" },
KEY(bytes_in), 4096, 30, 40, 200, 300, g_float, "bytes/sec", "%.2f" },
KEY(pkts_in), 256, 30, 40, 200, 300, g_float, "packets/sec", "%.2f" },
KEY(pkts_out), 256, 30, 40, 200, 300, g_float, "packets/sec", "%.2f" },
/*
* The amount of disk space could change - hot-swap, mounts, etc.
* check: 30-60min.
*/
KEY(disk_total), 1, 1800, 3600, 900, 1200, g_double, "GB", "%.3f" },
KEY(disk_free), 1, 30, 40, 120, 180, g_double, "GB", "%.3f" },
KEY(part_max_used), 1, 30, 40, 120, 180, g_float, "%", "%.1f" }
#endif
};
#endif /* METRIC_H */
