Skip to contentSkip to conversation

Recent changes (master)

PATCH Jens Axboe 1 participant 1 message Sep 21, 2018

lore
Share & tools
lore

Jens Axboe

author
#0 lore raw

The following changes since commit 5fa4aff59e4eec471b2c534702f0162006845c4b:

Merge branch 'epoch-time-hist-logs' of https://github.com/parallel-fs-utils/fio (2018-09-19 09:42:47 -0600)

are available in the git repository at:

git://git.kernel.dk/fio.git master

for you to fetch changes up to 4d2707ef027cc5b2418ab5de622318e0f70c096a:

blktrace: fix leak of 'merge_buf' (2018-09-20 14:25:56 -0600)

----------------------------------------------------------------
Dennis Zhou (4):
options: rename name string operations for more general use
blktrace: add support to interleave blktrace files
blktrace: add option to scale a trace
blktrace: add option to iterate over a trace multiple times

Jens Axboe (2):
blktrace: provide empty merge_blktrace_iologs()
blktrace: fix leak of 'merge_buf'

 HOWTO            |  71 +++++++++++++++++++
 blktrace.c       | 211 +++++++++++++++++++++++++++++++++++++++++++++++++++++++
 blktrace.h       |  22 ++++++
 cconv.c          |  15 ++++
 fio.1            |  66 +++++++++++++++++
 init.c           |  21 ++++++
 options.c        |  47 ++++++++++---
 options.h        |   2 +
 server.h         |   2 +-
 thread_options.h |   6 ++
 10 files changed, 453 insertions(+), 10 deletions(-)

---

Diff of recent changes:

HOWTO +71 -0
index 0c5b710..45cf0bd 100644
@@ -100,6 +100,10 @@ Command line options
 
 	Parse options only, don't start any I/O.
 
+.. option:: --merge-blktrace-only
+
+	Merge blktraces only, don't start any I/O.
+
 .. option:: --output=filename
 
 	Write output to file `filename`.
@@ -2491,6 +2495,33 @@ I/O replay
 	will be read at once. If selected true, input from iolog will be read
 	gradually. Useful when iolog is very large, or it is generated.
 
+.. option:: merge_blktrace_file=str
+
+	When specified, rather than replaying the logs passed to :option:`read_iolog`,
+	the logs go through a merge phase which aggregates them into a single
+	blktrace. The resulting file is then passed on as the :option:`read_iolog`
+	parameter. The intention here is to make the order of events consistent.
+	This limits the influence of the scheduler compared to replaying multiple
+	blktraces via concurrent jobs.
+
+.. option:: merge_blktrace_scalars=float_list
+
+	This is a percentage based option that is index paired with the list of
+	files passed to :option:`read_iolog`. When merging is performed, scale
+	the time of each event by the corresponding amount. For example,
+	``--merge_blktrace_scalars="50:100"`` runs the first trace in halftime
+	and the second trace in realtime. This knob is separately tunable from
+	:option:`replay_time_scale` which scales the trace during runtime and
+	does not change the output of the merge unlike this option.
+
+.. option:: merge_blktrace_iters=float_list
+
+	This is a whole number option that is index paired with the list of files
+	passed to :option:`read_iolog`. When merging is performed, run each trace
+	for the specified number of iterations. For example,
+	``--merge_blktrace_iters="2:1"`` runs the first trace for two iterations
+	and the second trace for one iteration.
+
 .. option:: replay_no_stall=bool
 
 	When replaying I/O with :option:`read_iolog` the default behavior is to
@@ -3839,6 +3870,46 @@ given in bytes. The `action` can be one of these:
 **trim**
 	   Trim the given file from the given `offset` for `length` bytes.
 
+
+I/O Replay - Merging Traces
+---------------------------
+
+Colocation is a common practice used to get the most out of a machine.
+Knowing which workloads play nicely with each other and which ones don't is
+a much harder task. While fio can replay workloads concurrently via multiple
+jobs, it leaves some variability up to the scheduler making results harder to
+reproduce. Merging is a way to make the order of events consistent.
+
+Merging is integrated into I/O replay and done when a
+:option:`merge_blktrace_file` is specified. The list of files passed to
+:option:`read_iolog` go through the merge process and output a single file
+stored to the specified file. The output file is passed on as if it were the
+only file passed to :option:`read_iolog`. An example would look like::
+
+	$ fio --read_iolog="<file1>:<file2>" --merge_blktrace_file="<output_file>"
+
+Creating only the merged file can be done by passing the command line argument
+:option:`merge-blktrace-only`.
+
+Scaling traces can be done to see the relative impact of any particular trace
+being slowed down or sped up. :option:`merge_blktrace_scalars` takes in a colon
+separated list of percentage scalars. It is index paired with the files passed
+to :option:`read_iolog`.
+
+With scaling, it may be desirable to match the running time of all traces.
+This can be done with :option:`merge_blktrace_iters`. It is index paired with
+:option:`read_iolog` just like :option:`merge_blktrace_scalars`.
+
+In an example, given two traces, A and B, each 60s long. If we want to see
+the impact of trace A issuing IOs twice as fast and repeat trace A over the
+runtime of trace B, the following can be done::
+
+	$ fio --read_iolog="<trace_a>:"<trace_b>" --merge_blktrace_file"<output_file>" --merge_blktrace_scalars="50:100" --merge_blktrace_iters="2:1"
+
+This runs trace A at 2x the speed twice for approximately the same runtime as
+a single run of trace B.
+
+
 CPU idleness profiling
 ----------------------
 
blktrace.c +211 -0

220 lines not on this page · raw message

blktrace.h +22 -0
index 096993e..a0e82fa 100644
@@ -1,10 +1,27 @@
 #ifndef FIO_BLKTRACE_H
 #define FIO_BLKTRACE_H
 
+
 #ifdef FIO_HAVE_BLKTRACE
 
+#include <asm/types.h>
+
+#include "blktrace_api.h"
+
+struct blktrace_cursor {
+	struct fifo		*fifo;	// fifo queue for reading
+	int			fd;	// blktrace file
+	__u64			length; // length of trace
+	struct blk_io_trace	t;	// current io trace
+	int			swap;	// bitwise reverse required
+	int			scalar;	// scale percentage
+	int			iter;	// current iteration
+	int			nr_iter; // number of iterations to run
+};
+
 bool is_blktrace(const char *, int *);
 bool load_blktrace(struct thread_data *, const char *, int);
+int merge_blktrace_iologs(struct thread_data *td);
 
 #else
 
@@ -19,5 +36,10 @@ static inline bool load_blktrace(struct thread_data *td, const char *fname,
 	return false;
 }
 
+static inline int merge_blktrace_iologs(struct thread_data *td)
+{
+	return false;
+}
+
 #endif
 #endif
cconv.c +15 -0
index 1d7f6f2..50e45c6 100644
@@ -37,6 +37,7 @@ static void free_thread_options_to_cpu(struct thread_options *o)
 	free(o->mmapfile);
 	free(o->read_iolog_file);
 	free(o->write_iolog_file);
+	free(o->merge_blktrace_file);
 	free(o->bw_log_file);
 	free(o->lat_log_file);
 	free(o->iops_log_file);
@@ -73,6 +74,7 @@ void convert_thread_options_to_cpu(struct thread_options *o,
 	string_to_cpu(&o->mmapfile, top->mmapfile);
 	string_to_cpu(&o->read_iolog_file, top->read_iolog_file);
 	string_to_cpu(&o->write_iolog_file, top->write_iolog_file);
+	string_to_cpu(&o->merge_blktrace_file, top->merge_blktrace_file);
 	string_to_cpu(&o->bw_log_file, top->bw_log_file);
 	string_to_cpu(&o->lat_log_file, top->lat_log_file);
 	string_to_cpu(&o->iops_log_file, top->iops_log_file);
@@ -304,6 +306,12 @@ void convert_thread_options_to_cpu(struct thread_options *o,
 
 	for (i = 0; i < FIO_IO_U_LIST_MAX_LEN; i++)
 		o->percentile_list[i].u.f = fio_uint64_to_double(le64_to_cpu(top->percentile_list[i].u.i));
+
+	for (i = 0; i < FIO_IO_U_LIST_MAX_LEN; i++)
+		o->merge_blktrace_scalars[i].u.f = fio_uint64_to_double(le64_to_cpu(top->merge_blktrace_scalars[i].u.i));
+
+	for (i = 0; i < FIO_IO_U_LIST_MAX_LEN; i++)
+		o->merge_blktrace_iters[i].u.f = fio_uint64_to_double(le64_to_cpu(top->merge_blktrace_iters[i].u.i));
 #if 0
 	uint8_t cpumask[FIO_TOP_STR_MAX];
 	uint8_t verify_cpumask[FIO_TOP_STR_MAX];
@@ -330,6 +338,7 @@ void convert_thread_options_to_net(struct thread_options_pack *top,
 	string_to_net(top->mmapfile, o->mmapfile);
 	string_to_net(top->read_iolog_file, o->read_iolog_file);
 	string_to_net(top->write_iolog_file, o->write_iolog_file);
+	string_to_net(top->merge_blktrace_file, o->merge_blktrace_file);
 	string_to_net(top->bw_log_file, o->bw_log_file);
 	string_to_net(top->lat_log_file, o->lat_log_file);
 	string_to_net(top->iops_log_file, o->iops_log_file);
@@ -565,6 +574,12 @@ void convert_thread_options_to_net(struct thread_options_pack *top,
 
 	for (i = 0; i < FIO_IO_U_LIST_MAX_LEN; i++)
 		top->percentile_list[i].u.i = __cpu_to_le64(fio_double_to_uint64(o->percentile_list[i].u.f));
+
+	for (i = 0; i < FIO_IO_U_LIST_MAX_LEN; i++)
+		top->merge_blktrace_scalars[i].u.i = __cpu_to_le64(fio_double_to_uint64(o->merge_blktrace_scalars[i].u.f));
+
+	for (i = 0; i < FIO_IO_U_LIST_MAX_LEN; i++)
+		top->merge_blktrace_iters[i].u.i = __cpu_to_le64(fio_double_to_uint64(o->merge_blktrace_iters[i].u.f));
 #if 0
 	uint8_t cpumask[FIO_TOP_STR_MAX];
 	uint8_t verify_cpumask[FIO_TOP_STR_MAX];
fio.1 +66 -0

84 lines not on this page · raw message

init.c +21 -0
index c235b05..560da8f 100644
@@ -30,6 +30,7 @@
 #include "idletime.h"
 #include "filelock.h"
 #include "steadystate.h"
+#include "blktrace.h"
 
 #include "oslib/getopt.h"
 #include "oslib/strcasestr.h"
@@ -46,6 +47,7 @@ static char **ini_file;
 static int max_jobs = FIO_MAX_JOBS;
 static int dump_cmdline;
 static int parse_only;
+static int merge_blktrace_only;
 
 static struct thread_data def_thread;
 struct thread_data *threads = NULL;
@@ -287,6 +289,11 @@ static struct option l_opts[FIO_NR_OPTIONS] = {
 		.val		= 'K',
 	},
 	{
+		.name		= (char *) "merge-blktrace-only",
+		.has_arg	= no_argument,
+		.val		= 'A' | FIO_CLIENT_FLAG,
+	},
+	{
 		.name		= NULL,
 	},
 };
@@ -1724,6 +1731,14 @@ static int add_job(struct thread_data *td, const char *jobname, int job_add_num,
 	if (td_steadystate_init(td))
 		goto err;
 
+	if (o->merge_blktrace_file && !merge_blktrace_iologs(td))
+		goto err;
+
+	if (merge_blktrace_only) {
+		put_job(td);
+		return 0;
+	}
+
 	/*
 	 * recurse add identical jobs, clear numjobs and stonewall options
 	 * as they don't apply to sub-jobs
@@ -2173,6 +2188,7 @@ static void usage(const char *name)
 	printf("  --debug=options\tEnable debug logging. May be one/more of:\n");
 	show_debug_categories();
 	printf("  --parse-only\t\tParse options only, don't start any IO\n");
+	printf("  --merge-blktrace-only\tMerge blktraces only, don't start any IO\n");
 	printf("  --output\t\tWrite output to file\n");
 	printf("  --bandwidth-log\tGenerate aggregate bandwidth logs\n");
 	printf("  --minimal\t\tMinimal (terse) output\n");
@@ -2889,6 +2905,11 @@ int parse_cmd_line(int argc, char *argv[], int client_type)
 			}
 			trigger_timeout /= 1000000;
 			break;
+
+		case 'A':
+			did_arg = true;
+			merge_blktrace_only = 1;
+			break;
 		case '?':
 			log_err("%s: unrecognized option '%s'\n", argv[0],
 							argv[optind - 1]);
options.c +38 -9

97 lines not on this page · raw message

options.h +2 -0
index 8fdd136..5276f31 100644
@@ -16,6 +16,8 @@ void add_opt_posval(const char *, const char *, const char *);
 void del_opt_posval(const char *, const char *);
 struct thread_data;
 void fio_options_free(struct thread_data *);
+char *get_next_str(char **ptr);
+int get_max_str_idx(char *input);
 char* get_name_by_idx(char *input, int index);
 int set_name_idx(char *, size_t, char *, int, bool);
 
server.h +1 -1
index 37d2f76..40b9eac 100644
@@ -48,7 +48,7 @@ struct fio_net_cmd_reply {
 };
 
 enum {
-	FIO_SERVER_VER			= 74,
+	FIO_SERVER_VER			= 77,
 
 	FIO_SERVER_MAX_FRAGMENT_PDU	= 1024,
 	FIO_SERVER_MAX_CMD_MB		= 2048,
thread_options.h +6 -0
index 3931583..4f791cf 100644
@@ -258,6 +258,9 @@ struct thread_options {
 	char *read_iolog_file;
 	bool read_iolog_chunked;
 	char *write_iolog_file;
+	char *merge_blktrace_file;
+	fio_fp64_t merge_blktrace_scalars[FIO_IO_U_LIST_MAX_LEN];
+	fio_fp64_t merge_blktrace_iters[FIO_IO_U_LIST_MAX_LEN];
 
 	unsigned int write_bw_log;
 	unsigned int write_lat_log;
@@ -540,6 +543,9 @@ struct thread_options_pack {
 
 	uint8_t read_iolog_file[FIO_TOP_STR_MAX];
 	uint8_t write_iolog_file[FIO_TOP_STR_MAX];
+	uint8_t merge_blktrace_file[FIO_TOP_STR_MAX];
+	fio_fp64_t merge_blktrace_scalars[FIO_IO_U_LIST_MAX_LEN];
+	fio_fp64_t merge_blktrace_iters[FIO_IO_U_LIST_MAX_LEN];
 
 	uint32_t write_bw_log;
 	uint32_t write_lat_log;