Dne 31.08.2012 00:01, Tomas Vondra napsal:
On 30 Srpen 2012, 23:44, Robert Haas wrote:
On Thu, Aug 30, 2012 at 3:48 PM, Tomas Vondra <t...@fuzzy.cz> wrote:
That sounds like a pretty trivial patch. I've been thinking about
yet
another option - histograms (regular or with exponential bins).
I thought about that, too, but I think high-outliers is a lot more
useful. At least for the kinds of things I worry about.
OK, let's fix the current patches first. We can add more options
later.
So, here is a fixed version of the patch. I've made these changes:
1) The option is now '--sampling-rate' instead of '-R' and accepts
float arguments. I've decided not to use 0.01 = 1% but use 1 = 1%, so it
accepts values between 0 and 100. I find this more natural.
2) I've removed one of the two new variables, the remaining one is used
both to check whether the sampling is enabled and keep the value.
3) I've decided not to enable "-l" whenever the "--sampling-rate" is
used. Again, I find this a bit less magic behavior.
4) I've fixed some formatting issues - if you notice another one that I
don't see, let me know.
Tomas
diff --git a/contrib/pgbench/pgbench.c b/contrib/pgbench/pgbench.c
index 00cab73..578aeb3 100644
--- a/contrib/pgbench/pgbench.c
+++ b/contrib/pgbench/pgbench.c
@@ -130,6 +130,11 @@ int foreign_keys = 0;
int unlogged_tables = 0;
/*
+ * use log sampling (rate => 1 = 100%, 0 = don't use)
+ */
+double use_sample_rate = 0.0;
+
+/*
* tablespace selection
*/
char *tablespace = NULL;
@@ -364,6 +369,8 @@ usage(void)
" -f FILENAME read transaction script from FILENAME\n"
" -j NUM number of threads (default: 1)\n"
" -l write transaction times to log file\n"
+ " --sampling-rate NUM\n"
+ " sampling rate of the log (e.g. 0.01 for 1%%
sample)\n"
" -M simple|extended|prepared\n"
" protocol for submitting queries to server
(default: simple)\n"
" -n do not run VACUUM before tests\n"
@@ -877,21 +884,26 @@ top:
instr_time diff;
double usec;
- INSTR_TIME_SET_CURRENT(now);
- diff = now;
- INSTR_TIME_SUBTRACT(diff, st->txn_begin);
- usec = (double) INSTR_TIME_GET_MICROSEC(diff);
+ /* either no sampling or is within the sample */
+ if ((use_sample_rate == 0.0) ||
(pg_erand48(thread->random_state) <= use_sample_rate))
+ {
+
+ INSTR_TIME_SET_CURRENT(now);
+ diff = now;
+ INSTR_TIME_SUBTRACT(diff, st->txn_begin);
+ usec = (double) INSTR_TIME_GET_MICROSEC(diff);
#ifndef WIN32
- /* This is more than we really ought to know about
instr_time */
- fprintf(logfile, "%d %d %.0f %d %ld %ld\n",
- st->id, st->cnt, usec, st->use_file,
- (long) now.tv_sec, (long) now.tv_usec);
+ /* This is more than we really ought to know
about instr_time */
+ fprintf(logfile, "%d %d %.0f %d %ld %ld\n",
+ st->id, st->cnt, usec,
st->use_file,
+ (long) now.tv_sec, (long)
now.tv_usec);
#else
- /* On Windows, instr_time doesn't provide a timestamp
anyway */
- fprintf(logfile, "%d %d %.0f %d 0 0\n",
- st->id, st->cnt, usec, st->use_file);
+ /* On Windows, instr_time doesn't provide a
timestamp anyway */
+ fprintf(logfile, "%d %d %.0f %d 0 0\n",
+ st->id, st->cnt, usec,
st->use_file);
#endif
+ }
}
if (commands[st->state]->type == SQL_COMMAND)
@@ -1918,6 +1930,7 @@ main(int argc, char **argv)
{"index-tablespace", required_argument, NULL, 3},
{"tablespace", required_argument, NULL, 2},
{"unlogged-tables", no_argument, &unlogged_tables, 1},
+ {"sampling-rate", required_argument, NULL, 4},
{NULL, 0, NULL, 0}
};
@@ -2123,6 +2136,14 @@ main(int argc, char **argv)
case 3: /* index-tablespace */
index_tablespace = optarg;
break;
+ case 4:
+ use_sample_rate = atof(optarg)/100;
+ if (use_sample_rate <= 0.0 || use_sample_rate >
1.0)
+ {
+ fprintf(stderr, "invalid sampling rate:
%f\n", use_sample_rate);
+ exit(1);
+ }
+ break;
default:
fprintf(stderr, _("Try \"%s --help\" for more
information.\n"), progname);
exit(1);
@@ -2158,6 +2179,12 @@ main(int argc, char **argv)
exit(1);
}
+ /* --sampling-rate may be used only with -l */
+ if (use_sample_rate > 0 && (! use_log)) {
+ fprintf(stderr, "log sampling rate is allowed only when logging
transactions\n");
+ exit(1);
+ }
+
/*
* is_latencies only works with multiple threads in thread-based
* implementations, not fork-based ones, because it supposes that the
diff --git a/doc/src/sgml/pgbench.sgml b/doc/src/sgml/pgbench.sgml
index 437fcea..67b21aa 100644
--- a/doc/src/sgml/pgbench.sgml
+++ b/doc/src/sgml/pgbench.sgml
@@ -317,6 +317,22 @@ pgbench <optional> <replaceable>options</> </optional>
<replaceable>dbname</>
</varlistentry>
<varlistentry>
+ <term><option>--sampling-rate</option> <replaceable>rate</></term>
+ <listitem>
+ <para>
+ Sampling rate, used when writing data into the log in percent. 100
means all
+ transactions will be logged, 1 means only 1% of the transactions will
be logged.
+ This may also be a float, e.g. 12.5 means 12.5% will be sampled.
+ </para>
+ <para>
+ Be careful when processing the log file - e.g. when computing tps
values, you
+ need to multiply the numbers accordingly (e.g. with 1% sample you'll
get 1/100
+ of the actual tps).
+ </para>
+ </listitem>
+ </varlistentry>
+
+ <varlistentry>
<term><option>-M</option> <replaceable>querymode</></term>
<listitem>
<para>
--
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers
