=head1 VALUE IDENTIFIER
-The values in the collectd are identified using an five-tupel (host, plugin,
+The values in the collectd are identified using an five-tuple (host, plugin,
plugin-instance, type, type-instance) where only plugin-instance and
type-instance may be NULL (or undefined). Many functions expect an
I<%identifier> hash that has at least the members B<host>, B<plugin>, and
=item I<$obj>-E<gt>B<putval> (I<%identifier>, B<time> =E<gt> I<$time>, B<values> =E<gt> [...]);
Submits a value-list to the daemon. If the B<time> argument is omitted
-C<time()> is used. The requierd argument B<values> is a reference to an array
+C<time()> is used. The required argument B<values> is a reference to an array
of values that is to be submitted. The number of values must match the number
of values expected for the given B<type> (see L<VALUE IDENTIFIER>), though this
is checked by the daemon, not the Perl module. Also, gauge data-sources
#!/usr/bin/perl
#
# collectd - contrib/cussh.pl
-# Copyright (C) 2007 Sebastian Harl
+# Copyright (C) 2007-2008 Sebastian Harl
#
# This program is free software; you can redistribute it and/or modify it
# under the terms of the GNU General Public License as published by the
my $sock = Collectd::Unixsock->new($path);
my $cmds = {
- PUTVAL => \&putval,
- GETVAL => \&getval,
- FLUSH => \&flush,
+ PUTVAL => \&putval,
+ GETVAL => \&getval,
+ FLUSH => \&flush,
+ LISTVAL => \&listval,
};
if (! $sock) {
exit 1;
}
- print "cussh version 0.2, Copyright (C) 2007 Sebastian Harl\n"
+ print "cussh version 0.2, Copyright (C) 2007-2008 Sebastian Harl\n"
. "cussh comes with ABSOLUTELY NO WARRANTY. This is free software,\n"
. "and you are welcome to redistribute it under certain conditions.\n"
. "See the GNU General Public License 2 for more details.\n\n";
print "cussh> ";
my $line = <STDIN>;
- last if ((! $line) || ($line =~ m/^quit$/i));
+ last if (! $line);
- my ($cmd) = $line =~ m/^(\w+)\s+/;
+ chomp $line;
+
+ last if ($line =~ m/^quit$/i);
+
+ my ($cmd) = $line =~ m/^(\w+)\s*/;
$line = $';
next if (! $cmd);
print $$string . $/;
my ($h, $p, $pi, $t, $ti) =
- $$string =~ m/^(\w+)\/(\w+)(?:-(\w+))?\/(\w+)(?:-(\w+))?\s+/;
+ $$string =~ m/^(\w+)\/(\w+)(?:-(\w+))?\/(\w+)(?:-(\w+))?\s*/;
$$string = $';
return if ((! $h) || (! $p) || (! $t));
return \%id;
}
+sub putid {
+ my $ident = shift || return;
+
+ my $string;
+
+ $string = $ident->{'host'} . "/" . $ident->{'plugin'};
+
+ if (defined $ident->{'plugin_instance'}) {
+ $string .= "-" . $ident->{'plugin_instance'};
+ }
+
+ $string .= "/" . $ident->{'type'};
+
+ if (defined $ident->{'type_instance'}) {
+ $string .= "-" . $ident->{'type_instance'};
+ }
+ return $string;
+}
+
=head1 COMMANDS
=over 4
=item B<GETVAL> I<Identifier>
-=item B<PUTVAL> I<Identifier> I<Valuelist>
-
-These commands follow the exact same syntax as described in
-L<collectd-unixsock(5)>.
-
=cut
sub putval {
}
my ($time, @values) = split m/:/, $line;
- return $sock->putval(%$id, $time, \@values);
+ return $sock->putval(%$id, time => $time, values => \@values);
}
+=item B<PUTVAL> I<Identifier> I<Valuelist>
+
+=cut
+
sub getval {
my $sock = shift || return;
my $line = shift || return;
my $vals = $sock->getval(%$id);
- return if (! $vals);
+ if (! $vals) {
+ print STDERR $sock->{'error'} . $/;
+ return;
+ }
foreach my $key (keys %$vals) {
print "\t$key: $vals->{$key}\n";
return 1;
}
+=item B<FLUSH> [B<timeout>=I<$timeout>] [B<plugin>=I<$plugin>[ ...]]
+
+=cut
+
sub flush {
my $sock = shift || return;
my $line = shift;
return 1;
}
+=item B<LISTVAL>
+
+=cut
+
+sub listval {
+ my $sock = shift || return;
+
+ my @res;
+
+ @res = $sock->listval();
+
+ if (! @res) {
+ print STDERR $sock->{'error'} . $/;
+ return;
+ }
+
+ foreach my $ident (@res) {
+ print $ident->{'time'} . " " . putid($ident) . $/;
+ }
+ return 1;
+}
+
+=back
+
+These commands follow the exact same syntax as described in
+L<collectd-unixsock(5)>.
+
=head1 SEE ALSO
L<collectd(1)>, L<collectd-unisock(5)>
connections. Once a connection is established the client can send commands to
the daemon which it will answer, if it understand them.
+In general the plugin answers with a status line of the following form:
+
+I<Status> I<Message>
+
+If I<Status> is greater than or equal to zero the message indicates success,
+if I<Status> is less than zero the message indicates failure. I<Message> is a
+human-readable string that further describes the return value.
+
+On success, I<Status> furthermore indicates the number of subsequent lines of
+output (not including the status line). Each such lines usually contains a
+single return value. See the description of each command for details.
+
The following commands are implemented:
=over 4
=item B<GETVAL> I<Identifier>
If the value identified by I<Identifier> (see below) is found the complete
-value-list is returned. The response is a space separated list of
-name-value-pairs:
-
-I<num> I<name>B<=>I<value>[ I<name>B<=>I<value>[ ...]]
-
-If I<num> is less then zero, an error occurred. Otherwise it contains the
-number of values that follow. Each value is of the form I<name>B<=>I<value>.
+value-list is returned. The response is a list of name-value-pairs, each pair
+on its own line (the number of lines is indicated by the status line - see
+above). Each name-value-pair is of the form I<name>B<=>I<value>.
Counter-values are converted to a rate, e.E<nbsp>g. bytes per second.
Undefined values are returned as B<NaN>.
Example:
-> | GETVAL myhost/cpu-0/cpu-user
- <- | 1 value=1.260000e+00
+ <- | 1 Value found
+ <- | value=1.260000e+00
=item B<LISTVAL>
Returns a list of the values available in the value cache together with the
time of the last update, so that querying applications can issue a B<GETVAL>
-command for the values that have changed.
-
-The first line's status number is the number of identifiers returned or less
-than zero if an error occurred. Each of the following lines contains the
+command for the values that have changed. Each return value consists of the
update time as an epoch value and the identifier, separated by a space. The
update time is the time of the last value, as provided by the collecting
-instance and may be very different from the time the server consideres to be
+instance and may be very different from the time the server considers to be
"now".
Example:
-> | LISTVAL
<- | 69 Values found
- <- | 1182204284 leeloo/cpu-0/cpu-idle
- <- | 1182204284 leeloo/cpu-0/cpu-nice
- <- | 1182204284 leeloo/cpu-0/cpu-system
- <- | 1182204284 leeloo/cpu-0/cpu-user
+ <- | 1182204284 myhost/cpu-0/cpu-idle
+ <- | 1182204284 myhost/cpu-0/cpu-nice
+ <- | 1182204284 myhost/cpu-0/cpu-system
+ <- | 1182204284 myhost/cpu-0/cpu-user
...
=item B<PUTVAL> I<Identifier> [I<OptionList>] I<Valuelist>
=item B<FLUSH> [B<timeout=>I<Timeout>] [B<plugin=>I<Plugin> [...]]
Flushes all cached data older than I<Timeout> seconds. If no timeout has been
-specified, it defaults to -1 which causes all data to be flushed.
+specified, it defaults to -1 which causes all data to be flushed. B<timeout>
+may be specified multiple times - each occurrence applies to plugins listed
+afterwards.
If specified, only specific plugins are flushed. Otherwise all plugins
providing a flush callback are flushed.
myhost/memory/memory-used
myhost/disk-sda/disk_octets
-=head2 Return values
-
-Unless otherwise noted the plugin answers with a line of the following form:
-
-I<Num> I<Message>
-
-If I<Num> is zero the message indicates success, if I<Num> is non-zero the
-message indicates failure. I<Message> is a human-readable string that describes
-the return value further.
-
-Commands that return values may use I<Num> to return the number of values that
-follow, such as the B<GETVAL> command. These commands usually return a negative
-value on failure and never return zero.
-
=head1 ABSTRACTION LAYER
-Shipped with the sourcecode comes the Perl-Module L<Collectd::Unixsock> which
+B<collectd> ships the Perl-Module L<Collectd::Unixsock> which
provides an abstraction layer over the actual socket connection. It can be
-found in the directory F<contrib/PerlLib>. If you want to use Perl to
-communicate with the daemon, you're encouraged to use and expand this module.
+found in the directory F<bindings/perl/> in the source distribution or
+(usually) somewhere near F</usr/share/perl5/> if you're using a package. If
+you want to use Perl to communicate with the daemon, you're encouraged to use
+and expand this module.
=head1 SEE ALSO
static void *us_handle_client (void *arg)
{
int fd;
- FILE *fh;
+ FILE *fhin, *fhout;
char buffer[1024];
char *fields[128];
int fields_num;
DEBUG ("Reading from fd #%i", fd);
- fh = fdopen (fd, "r+");
- if (fh == NULL)
+ fhin = fdopen (fd, "r");
+ if (fhin == NULL)
{
char errbuf[1024];
ERROR ("unixsock plugin: fdopen failed: %s",
pthread_exit ((void *) 1);
}
- while (fgets (buffer, sizeof (buffer), fh) != NULL)
+ fhout = fdopen (fd, "w");
+ if (fhout == NULL)
+ {
+ char errbuf[1024];
+ ERROR ("unixsock plugin: fdopen failed: %s",
+ sstrerror (errno, errbuf, sizeof (errbuf)));
+ fclose (fhin); /* this closes fd as well */
+ pthread_exit ((void *) 1);
+ }
+
+ while (fgets (buffer, sizeof (buffer), fhin) != NULL)
{
int len;
if (strcasecmp (fields[0], "getval") == 0)
{
- handle_getval (fh, fields, fields_num);
+ handle_getval (fhout, fields, fields_num);
}
else if (strcasecmp (fields[0], "putval") == 0)
{
- handle_putval (fh, fields, fields_num);
+ handle_putval (fhout, fields, fields_num);
}
else if (strcasecmp (fields[0], "listval") == 0)
{
- handle_listval (fh, fields, fields_num);
+ handle_listval (fhout, fields, fields_num);
}
else if (strcasecmp (fields[0], "putnotif") == 0)
{
- handle_putnotif (fh, fields, fields_num);
+ handle_putnotif (fhout, fields, fields_num);
}
else if (strcasecmp (fields[0], "flush") == 0)
{
- handle_flush (fh, fields, fields_num);
+ handle_flush (fhout, fields, fields_num);
}
else
{
- fprintf (fh, "-1 Unknown command: %s\n", fields[0]);
- fflush (fh);
+ fprintf (fhout, "-1 Unknown command: %s\n", fields[0]);
+ fflush (fhout);
}
} /* while (fgets) */
DEBUG ("Exiting..");
- close (fd);
+ fclose (fhin);
+ fclose (fhout);
pthread_exit ((void *) 0);
return ((void *) 0);
#include "common.h"
#include "plugin.h"
-struct flush_info_s
+int handle_flush (FILE *fh, char **fields, int fields_num)
{
- char **plugins;
- int plugins_num;
- int timeout;
-};
-typedef struct flush_info_s flush_info_t;
+ int success = 0;
+ int error = 0;
-static int parse_option_plugin (flush_info_t *fi, const char *option)
-{
- char **temp;
+ int timeout = -1;
- temp = (char **) realloc (fi->plugins,
- (fi->plugins_num + 1) * sizeof (char *));
- if (temp == NULL)
- {
- ERROR ("utils_cmd_flush: parse_option_plugin: realloc failed.");
- return (-1);
- }
- fi->plugins = temp;
+ int i;
- fi->plugins[fi->plugins_num] = strdup (option + strlen ("plugin="));
- if (fi->plugins[fi->plugins_num] == NULL)
+ for (i = 1; i < fields_num; i++)
{
- /* fi->plugins is freed in handle_flush in this case */
- ERROR ("utils_cmd_flush: parse_option_plugin: strdup failed.");
- return (-1);
- }
- fi->plugins_num++;
+ char *option = fields[i];
+ int status = 0;
- return (0);
-} /* int parse_option_plugin */
-
-static int parse_option_timeout (flush_info_t *fi, const char *option)
-{
- const char *value_ptr = option + strlen ("timeout=");
- char *endptr = NULL;
- int timeout;
-
- timeout = strtol (value_ptr, &endptr, 0);
- if (value_ptr == endptr)
- return (-1);
-
- fi->timeout = (timeout <= 0) ? (-1) : timeout;
-
- return (0);
-} /* int parse_option_timeout */
+ if (strncasecmp ("plugin=", option, strlen ("plugin=")) == 0)
+ {
+ char *plugin = option + strlen ("plugin=");
-static int parse_option (flush_info_t *fi, const char *option)
-{
- if (strncasecmp ("plugin=", option, strlen ("plugin=")) == 0)
- return (parse_option_plugin (fi, option));
- else if (strncasecmp ("timeout=", option, strlen ("timeout=")) == 0)
- return (parse_option_timeout (fi, option));
- else
- return (-1);
-} /* int parse_option */
+ if (0 == plugin_flush_one (timeout, plugin))
+ ++success;
+ else
+ ++error;
+ }
+ else if (strncasecmp ("timeout=", option, strlen ("timeout=")) == 0)
+ {
+ char *endptr = NULL;
+ char *value = option + strlen ("timeout=");
-int handle_flush (FILE *fh, char **fields, int fields_num)
-{
- flush_info_t fi;
- int status;
- int i;
+ errno = 0;
+ timeout = strtol (value, &endptr, 0);
- memset (&fi, '\0', sizeof (fi));
- fi.timeout = -1;
+ if ((endptr == value) || (0 != errno))
+ status = -1;
+ else if (0 >= timeout)
+ timeout = -1;
+ }
+ else
+ status = -1;
- for (i = 1; i < fields_num; i++)
- {
- status = parse_option (&fi, fields[i]);
if (status != 0)
{
- fprintf (fh, "-1 Cannot parse option %s\n", fields[i]);
+ fprintf (fh, "-1 Cannot parse option %s\n", option);
fflush (fh);
return (-1);
}
}
- if (fi.plugins_num > 0)
+ if ((success + error) > 0)
{
- int success = 0;
- for (i = 0; i < fi.plugins_num; i++)
- {
- status = plugin_flush_one (fi.timeout, fi.plugins[i]);
- if (status == 0)
- success++;
- }
- fprintf (fh, "0 Done: %i successful, %i errors\n",
- success, fi.plugins_num - success);
+ fprintf (fh, "0 Done: %i successful, %i errors\n", success, error);
}
else
{
- plugin_flush_all (fi.timeout);
+ plugin_flush_all (timeout);
fprintf (fh, "0 Done\n");
}
fflush (fh);
- for (i = 0; i < fi.plugins_num; i++)
- {
- sfree (fi.plugins[i]);
- }
- sfree (fi.plugins);
-
return (0);
} /* int handle_flush */
gauge_t *values;
size_t values_num;
+ char *identifier_copy;
+
const data_set_t *ds;
int status;
if (strlen (fields[1]) < strlen ("h/p/t"))
{
- fprintf (fh, "-1 Invalied identifier, %s", fields[1]);
+ fprintf (fh, "-1 Invalied identifier, %s\n", fields[1]);
fflush (fh);
return (-1);
}
- status = parse_identifier (fields[1], &hostname,
+ /* parse_identifier() modifies its first argument,
+ * returning pointers into it */
+ identifier_copy = sstrdup (fields[1]);
+
+ status = parse_identifier (identifier_copy, &hostname,
&plugin, &plugin_instance,
&type, &type_instance);
if (status != 0)
DEBUG ("unixsock plugin: Cannot parse `%s'", fields[1]);
fprintf (fh, "-1 Cannot parse identifier.\n");
fflush (fh);
+ sfree (identifier_copy);
return (-1);
}
DEBUG ("unixsock plugin: plugin_get_ds (%s) == NULL;", type);
fprintf (fh, "-1 Type `%s' is unknown.\n", type);
fflush (fh);
+ sfree (identifier_copy);
return (-1);
}
status = uc_get_rate_by_name (fields[1], &values, &values_num);
if (status != 0)
{
- fprintf (fh, "-1 No such value");
+ fprintf (fh, "-1 No such value\n");
fflush (fh);
+ sfree (identifier_copy);
return (-1);
}
fprintf (fh, "-1 Error reading value from cache.\n");
fflush (fh);
sfree (values);
+ sfree (identifier_copy);
return (-1);
}
- fprintf (fh, "%u", (unsigned int) values_num);
+ fprintf (fh, "%u Value%s found\n", (unsigned int) values_num,
+ (values_num == 1) ? "" : "s");
for (i = 0; i < values_num; i++)
{
- fprintf (fh, " %s=", ds->ds[i].name);
+ fprintf (fh, "%s=", ds->ds[i].name);
if (isnan (values[i]))
- fprintf (fh, "NaN");
+ fprintf (fh, "NaN\n");
else
- fprintf (fh, "%12e", values[i]);
+ fprintf (fh, "%12e\n", values[i]);
}
-
- fprintf (fh, "\n");
fflush (fh);
sfree (values);
+ sfree (identifier_copy);
return (0);
} /* int handle_getval */
return (-1);
}
- fprintf (fh, "%i Values found\n", (int) number);
+ fprintf (fh, "%i Value%s found\n", (int) number, (number == 1) ? "" : "s");
for (i = 0; i < number; i++)
fprintf (fh, "%u %s\n", (unsigned int) times[i], names[i]);
fflush (fh);
status = parse_option (&n, fields[i]);
if (status != 0)
{
- fprintf (fh, "-1 Error parsing option `%s'", fields[i]);
+ fprintf (fh, "-1 Error parsing option `%s'\n", fields[i]);
break;
}
}
char *value_str = strchr (time_str, ':');
if (value_str == NULL)
{
- fprintf (fh, "-1 No time found.");
+ fprintf (fh, "-1 No time found.\n");
return (-1);
}
*value_str = '\0'; value_str++;
int status;
int i;
+ char *identifier_copy;
+
const data_set_t *ds;
value_list_t vl = VALUE_LIST_INIT;
return (-1);
}
- status = parse_identifier (fields[1], &hostname,
+ /* parse_identifier() modifies its first argument,
+ * returning pointers into it */
+ identifier_copy = sstrdup (fields[1]);
+
+ status = parse_identifier (identifier_copy, &hostname,
&plugin, &plugin_instance,
&type, &type_instance);
if (status != 0)
DEBUG ("cmd putval: Cannot parse `%s'", fields[1]);
fprintf (fh, "-1 Cannot parse identifier.\n");
fflush (fh);
+ sfree (identifier_copy);
return (-1);
}
|| ((type_instance != NULL)
&& (strlen (type_instance) >= sizeof (vl.type_instance))))
{
- fprintf (fh, "-1 Identifier too long.");
+ fprintf (fh, "-1 Identifier too long.\n");
+ fflush (fh);
+ sfree (identifier_copy);
return (-1);
}
strcpy (vl.type_instance, type_instance);
ds = plugin_get_ds (type);
- if (ds == NULL)
+ if (ds == NULL) {
+ sfree (identifier_copy);
return (-1);
+ }
vl.values_len = ds->ds_num;
vl.values = (value_t *) malloc (vl.values_len * sizeof (value_t));
if (vl.values == NULL)
{
- fprintf (fh, "-1 malloc failed.");
+ fprintf (fh, "-1 malloc failed.\n");
+ fflush (fh);
+ sfree (identifier_copy);
return (-1);
}
{
if (parse_option (&vl, fields[i]) != 0)
{
- fprintf (fh, "-1 Error parsing option `%s'",
+ fprintf (fh, "-1 Error parsing option `%s'\n",
fields[i]);
break;
}
fflush (fh);
sfree (vl.values);
+ sfree (identifier_copy);
return (0);
} /* int handle_putval */
projects / collectd.git / commitdiff