Add Python 3 rewrite of arc_summary.py

Add new script arc_summary3.py as a complete rewrite of the
arc_summary.py tool (see issue #6873)

Add new options:

        -g/--graph    - Display crude graphic representation
                        of ARC status and quit
        -r/--raw      - Print all available information as
                        minimally formatted list (for grep)
        -s/--section  - Print a single section. This
                        replaces -p/--page, which is kept for
                        backwards use but marked as
                        depreciated

Add new sections with information on ZIL and SPL. Notify user
if sections L2ARC and VDEV are skipped instead of failing
silently. Add warning that -p/--page option is depreciated.

Developed for Python 3.5.

Reviewed-by: Richard Laager <rlaager@wiktel.com>
Reviewed-by: Richard Elling <Richard.Elling@RichardElling.com>
Reviewed by: George Melikov <mail@gmelikov.ru>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Scot W. Stevenson <scot.stevenson@gmail.com>
Closes #6873 
Closes #6892

cmd/arc_summary/Makefile.am

@@ -1 +1 @@
-dist_bin_SCRIPTS = arc_summary.py
+dist_bin_SCRIPTS = arc_summary.py arc_summary3.py

cmd/arc_summary/arc_summary3.py

@@ -0,0 +1,856 @@
+#!/usr/bin/python3
+#
+# Copyright (c) 2008 Ben Rockwood <benr@cuddletech.com>,
+# Copyright (c) 2010 Martin Matuska <mm@FreeBSD.org>,
+# Copyright (c) 2010-2011 Jason J. Hellenthal <jhell@DataIX.net>,
+# Copyright (c) 2017 Scot W. Stevenson <scot.stevenson@gmail.com>
+# All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+#
+# 1. Redistributions of source code must retain the above copyright
+#    notice, this list of conditions and the following disclaimer.
+# 2. Redistributions in binary form must reproduce the above copyright
+#    notice, this list of conditions and the following disclaimer in the
+#    documentation and/or other materials provided with the distribution.
+#
+# THIS SOFTWARE IS PROVIDED BY AUTHOR AND CONTRIBUTORS ``AS IS'' AND
+# ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+# ARE DISCLAIMED.  IN NO EVENT SHALL AUTHOR OR CONTRIBUTORS BE LIABLE
+# FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+# DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+# OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+# HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+# LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+# OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+# SUCH DAMAGE.
+"""Print statistics on the ZFS ARC Cache and other information
+
+Provides basic information on the ARC, its efficiency, the L2ARC (if present),
+the Data Management Unit (DMU), Virtual Devices (VDEVs), and tunables. See
+the in-source documentation and code at
+https://github.com/zfsonlinux/zfs/blob/master/module/zfs/arc.c for details.
+The original introduction to arc_summary can be found at
+http://cuddletech.com/?p=454
+"""
+
+import argparse
+import os
+import subprocess
+import sys
+import time
+
+DECRIPTION = 'Print ARC and other statistics for ZFS on Linux'
+INDENT = ' '*8
+LINE_LENGTH = 72
+PROC_PATH = '/proc/spl/kstat/zfs/'
+SPL_PATH = '/sys/module/spl/parameters/'
+TUNABLES_PATH = '/sys/module/zfs/parameters/'
+DATE_FORMAT = '%a %b %d %H:%M:%S %Y'
+TITLE = 'ZFS Subsystem Report'
+
+SECTIONS = 'arc archits dmu l2arc spl tunables vdev zil'.split()
+SECTION_HELP = 'print info from one section ('+' '.join(SECTIONS)+')'
+
+# Tunables and SPL are handled separately because they come from
+# different sources
+SECTION_PATHS = {'arc': 'arcstats',
+                 'dmu': 'dmu_tx',
+                 'l2arc': 'arcstats',  # L2ARC stuff lives in arcstats
+                 'vdev': 'vdev_cache_stats',
+                 'xuio': 'xuio_stats',
+                 'zfetch': 'zfetchstats',
+                 'zil': 'zil'}
+
+parser = argparse.ArgumentParser(description=DECRIPTION)
+parser.add_argument('-a', '--alternate', action='store_true', default=False,
+                    help='use alternate formatting for tunables and SPL',
+                    dest='alt')
+parser.add_argument('-d', '--description', action='store_true', default=False,
+                    help='print descriptions with tunables and SPL',
+                    dest='desc')
+parser.add_argument('-g', '--graph', action='store_true', default=False,
+                    help='print graph on ARC use and exit', dest='graph')
+parser.add_argument('-p', '--page', type=int, dest='page',
+                    help='print page by number (DEPRECATED, use "-s")')
+parser.add_argument('-r', '--raw', action='store_true', default=False,
+                    help='dump all available data with minimal formatting',
+                    dest='raw')
+parser.add_argument('-s', '--section', dest='section', help=SECTION_HELP)
+ARGS = parser.parse_args()
+
+
+def cleanup_line(single_line):
+    """Format a raw line of data from /proc and isolate the name value
+    part, returning a tuple with each. Currently, this gets rid of the
+    middle '4'. For example "arc_no_grow    4    0" returns the tuple
+    ("arc_no_grow", "0").
+    """
+    name, _, value = single_line.split()
+
+    return name, value
+
+
+def draw_graph(kstats_dict):
+    """Draw a primitive graph representing the basic information on the
+    ARC -- its size and the proportion used by MFU and MRU -- and quit.
+    We use max size of the ARC to calculate how full it is. This is a
+    very rough representation.
+    """
+
+    arc_stats = isolate_section('arcstats', kstats_dict)
+
+    GRAPH_INDENT = ' '*4
+    GRAPH_WIDTH = 60
+    arc_size = f_bytes(arc_stats['size'])
+    arc_perc = f_perc(arc_stats['size'], arc_stats['c_max'])
+    mfu_size = f_bytes(arc_stats['mfu_size'])
+    mru_size = f_bytes(arc_stats['mru_size'])
+
+    info_form = 'ARC: {0} ({1})  MFU: {2}  MRU: {3}'
+    info_line = info_form.format(arc_size, arc_perc, mfu_size, mru_size)
+    info_spc = ' '*int((GRAPH_WIDTH-len(info_line))/2)
+    info_line = GRAPH_INDENT+info_spc+info_line
+
+    graph_line = GRAPH_INDENT+'+'+('-'*(GRAPH_WIDTH-2))+'+'
+
+    mfu_perc = float(int(arc_stats['mfu_size'])/int(arc_stats['c_max']))
+    mru_perc = float(int(arc_stats['mru_size'])/int(arc_stats['c_max']))
+    arc_perc = float(int(arc_stats['size'])/int(arc_stats['c_max']))
+    total_ticks = float(arc_perc)*GRAPH_WIDTH
+    mfu_ticks = mfu_perc*GRAPH_WIDTH
+    mru_ticks = mru_perc*GRAPH_WIDTH
+    other_ticks = total_ticks-(mfu_ticks+mru_ticks)
+
+    core_form = 'F'*int(mfu_ticks)+'R'*int(mru_ticks)+'O'*int(other_ticks)
+    core_spc = ' '*(GRAPH_WIDTH-(2+len(core_form)))
+    core_line = GRAPH_INDENT+'|'+core_form+core_spc+'|'
+
+    for line in ('', info_line, graph_line, core_line, graph_line, ''):
+        print(line)
+
+
+def f_bytes(byte_string):
+    """Return human-readable representation of a byte value in
+    powers of 2 (eg "KiB" for "kibibytes", etc) to two decimal
+    points. Values smaller than one KiB are returned without
+    decimal points. Note "bytes" is a reserved keyword.
+    """
+
+    prefixes = ([2**80, "YiB"],   # yobibytes (yotta)
+                [2**70, "ZiB"],   # zebibytes (zetta)
+                [2**60, "EiB"],   # exbibytes (exa)
+                [2**50, "PiB"],   # pebibytes (peta)
+                [2**40, "TiB"],   # tebibytes (tera)
+                [2**30, "GiB"],   # gibibytes (giga)
+                [2**20, "MiB"],   # mebibytes (mega)
+                [2**10, "KiB"])   # kibibytes (kilo)
+
+    bites = int(byte_string)
+
+    if bites >= 2**10:
+        for limit, unit in prefixes:
+
+            if bites >= limit:
+                value = bites / limit
+                break
+
+        result = '{0:.1f} {1}'.format(value, unit)
+    else:
+        result = '{0} Bytes'.format(bites)
+
+    return result
+
+
+def f_hits(hits_string):
+    """Create a human-readable representation of the number of hits.
+    The single-letter symbols used are SI to avoid the confusion caused
+    by the different "short scale" and "long scale" representations in
+    English, which use the same words for different values. See
+    https://en.wikipedia.org/wiki/Names_of_large_numbers and:
+    https://physics.nist.gov/cuu/Units/prefixes.html
+    """
+
+    numbers = ([10**24, 'Y'],  # yotta (septillion)
+               [10**21, 'Z'],  # zetta (sextillion)
+               [10**18, 'E'],  # exa   (quintrillion)
+               [10**15, 'P'],  # peta  (quadrillion)
+               [10**12, 'T'],  # tera  (trillion)
+               [10**9, 'G'],   # giga  (billion)
+               [10**6, 'M'],   # mega  (million)
+               [10**3, 'k'])   # kilo  (thousand)
+
+    hits = int(hits_string)
+
+    if hits >= 1000:
+        for limit, symbol in numbers:
+
+            if hits >= limit:
+                value = hits/limit
+                break
+
+        result = "%0.1f%s" % (value, symbol)
+    else:
+        result = "%d" % hits
+
+    return result
+
+
+def f_perc(value1, value2):
+    """Calculate percentage and return in human-readable form. If
+    rounding produces the result '0.0' though the first number is
+    not zero, include a 'less-than' symbol to avoid confusion.
+    Division by zero is handled by returning 'n/a'; no error
+    is called.
+    """
+
+    v1 = float(value1)
+    v2 = float(value2)
+
+    try:
+        perc = 100 * v1/v2
+    except ZeroDivisionError:
+        result = 'n/a'
+    else:
+        result = '{0:0.1f} %'.format(perc)
+
+    if result == '0.0 %' and v1 > 0:
+        result = '< 0.1 %'
+
+    return result
+
+
+def format_raw_line(name, value):
+    """For the --raw option for the tunable and SPL outputs, decide on the
+    correct formatting based on the --alternate flag.
+    """
+
+    if ARGS.alt:
+        result = '{0}{1}={2}'.format(INDENT, name, value)
+    else:
+        spc = LINE_LENGTH-(len(INDENT)+len(value))
+        result = '{0}{1:<{spc}}{2}'.format(INDENT, name, value, spc=spc)
+
+    return result
+
+
+def get_kstats():
+    """Collect information on the ZFS subsystem from the /proc Linux virtual
+    file system. The step does not perform any further processing, giving us
+    the option to only work on what is actually needed. The name "kstat" is a
+    holdover from the Solaris utility of the same name.
+    """
+
+    result = {}
+    secs = SECTION_PATHS.values()
+
+    for section in secs:
+
+        with open(PROC_PATH+section, 'r') as proc_location:
+            lines = [line for line in proc_location]
+
+        del lines[0:2]  # Get rid of header
+        result[section] = lines
+
+    return result
+
+
+def get_spl_tunables(PATH):
+    """Collect information on the Solaris Porting Layer (SPL) or the
+    tunables, depending on the PATH given. Does not check if PATH is
+    legal.
+    """
+
+    result = {}
+    parameters = os.listdir(PATH)
+
+    for name in parameters:
+
+        with open(PATH+name, 'r') as para_file:
+            value = para_file.read()
+            result[name] = value.strip()
+
+    return result
+
+
+def get_descriptions(request):
+    """Get the decriptions of the Solaris Porting Layer (SPL) or the
+    tunables, return with minimal formatting.
+    """
+
+    if request not in ('spl', 'zfs'):
+        print('ERROR: description of "{0}" requested)'.format(request))
+        sys.exit(1)
+
+    descs = {}
+    target_prefix = 'parm:'
+
+    # We would prefer to do this with /sys/modules -- see the discussion at
+    # get_version() -- but there isn't a way to get the descriptions from
+    # there, so we fall back on modinfo
+    command = ["/sbin/modinfo", request, "-0"]
+
+    # The recommended way to do this is with subprocess.run(). However,
+    # some installed versions of Python are < 3.5, so we offer them
+    # the option of doing it the old way (for now)
+    info = ''
+
+    try:
+
+        if 'run' in dir(subprocess):
+            info = subprocess.run(command, stdout=subprocess.PIPE,
+                                  universal_newlines=True)
+            raw_output = info.stdout.split('\0')
+        else:
+            info = subprocess.check_output(command, universal_newlines=True)
+            raw_output = info.split('\0')
+
+    except subprocess.CalledProcessError:
+        print("Error: Descriptions not available (can't access kernel module)")
+        sys.exit(1)
+
+    for line in raw_output:
+
+        if not line.startswith(target_prefix):
+            continue
+
+        line = line[len(target_prefix):].strip()
+        name, raw_desc = line.split(':', 1)
+        desc = raw_desc.rsplit('(', 1)[0]
+
+        if desc == '':
+            desc = '(No description found)'
+
+        descs[name.strip()] = desc.strip()
+
+    return descs
+
+
+def get_version(request):
+    """Get the version number of ZFS or SPL on this machine for header.
+    Returns an error string, but does not raise an error, if we can't
+    get the ZFS/SPL version via modinfo.
+    """
+
+    if request not in ('spl', 'zfs'):
+        error_msg = '(ERROR: "{0}" requested)'.format(request)
+        return error_msg
+
+    # The original arc_summary.py called /sbin/modinfo/{spl,zfs} to get
+    # the version information. We switch to /sys/module/{spl,zfs}/version
+    # to make sure we get what is really loaded in the kernel
+    command = ["cat", "/sys/module/{0}/version".format(request)]
+    req = request.upper()
+    version = "(Can't get {0} version)".format(req)
+
+    # The recommended way to do this is with subprocess.run(). However,
+    # some installed versions of Python are < 3.5, so we offer them
+    # the option of doing it the old way (for now)
+    info = ''
+    if 'run' in dir(subprocess):
+        info = subprocess.run(command, stdout=subprocess.PIPE,
+                              universal_newlines=True)
+        version = info.stdout.strip()
+    else:
+        info = subprocess.check_output(command, universal_newlines=True)
+        version = info.strip()
+
+    return version
+
+
+def print_header():
+    """Print the initial heading with date and time as well as info on the
+    Linux and ZFS versions. This is not called for the graph.
+    """
+
+    # datetime is now recommended over time but we keep the exact formatting
+    # from the older version of arc_summary.py in case there are scripts
+    # that expect it in this way
+    daydate = time.strftime(DATE_FORMAT)
+    spc_date = LINE_LENGTH-len(daydate)
+    sys_version = os.uname()
+
+    sys_msg = sys_version.sysname+' '+sys_version.release
+    zfs = get_version('zfs')
+    spc_zfs = LINE_LENGTH-len(zfs)
+
+    machine_msg = 'Machine: '+sys_version.nodename+' ('+sys_version.machine+')'
+    spl = get_version('spl')
+    spc_spl = LINE_LENGTH-len(spl)
+
+    print('\n'+('-'*LINE_LENGTH))
+    print('{0:<{spc}}{1}'.format(TITLE, daydate, spc=spc_date))
+    print('{0:<{spc}}{1}'.format(sys_msg, zfs, spc=spc_zfs))
+    print('{0:<{spc}}{1}\n'.format(machine_msg, spl, spc=spc_spl))
+
+
+def print_raw(kstats_dict):
+    """Print all available data from the system in a minimally sorted format.
+    This can be used as a source to be piped through 'grep'.
+    """
+
+    sections = sorted(kstats_dict.keys())
+
+    for section in sections:
+
+        print('\n{0}:'.format(section.upper()))
+        lines = sorted(kstats_dict[section])
+
+        for line in lines:
+            name, value = cleanup_line(line)
+            print(format_raw_line(name, value))
+
+    # Tunables and SPL must be handled separately because they come from a
+    # different source and have descriptions the user might request
+    print()
+    section_spl()
+    section_tunables()
+
+
+def isolate_section(section_name, kstats_dict):
+    """From the complete information on all sections, retrieve only those
+    for one section.
+    """
+
+    try:
+        section_data = kstats_dict[section_name]
+    except KeyError:
+        print('ERROR: Data on {0} not available'.format(section_data))
+        sys.exit(1)
+
+    section_dict = dict(cleanup_line(l) for l in section_data)
+
+    return section_dict
+
+
+# Formatted output helper functions
+
+
+def prt_1(text, value):
+    """Print text and one value, no indent"""
+    spc = ' '*(LINE_LENGTH-(len(text)+len(value)))
+    print('{0}{spc}{1}'.format(text, value, spc=spc))
+
+
+def prt_i1(text, value):
+    """Print text and one value, with indent"""
+    spc = ' '*(LINE_LENGTH-(len(INDENT)+len(text)+len(value)))
+    print(INDENT+'{0}{spc}{1}'.format(text, value, spc=spc))
+
+
+def prt_2(text, value1, value2):
+    """Print text and two values, no indent"""
+    values = '{0:>9}  {1:>9}'.format(value1, value2)
+    spc = ' '*(LINE_LENGTH-(len(text)+len(values)+2))
+    print('{0}{spc}  {1}'.format(text, values, spc=spc))
+
+
+def prt_i2(text, value1, value2):
+    """Print text and two values, with indent"""
+    values = '{0:>9}  {1:>9}'.format(value1, value2)
+    spc = ' '*(LINE_LENGTH-(len(INDENT)+len(text)+len(values)+2))
+    print(INDENT+'{0}{spc}  {1}'.format(text, values, spc=spc))
+
+
+# The section output concentrates on important parameters instead of
+# being exhaustive (that is what the --raw parameter is for)
+
+
+def section_arc(kstats_dict):
+    """Give basic information on the ARC, MRU and MFU. This is the first
+    and most used section.
+    """
+
+    arc_stats = isolate_section('arcstats', kstats_dict)
+
+    throttle = arc_stats['memory_throttle_count']
+
+    if throttle == '0':
+        health = 'HEALTHY'
+    else:
+        health = 'THROTTLED'
+
+    prt_1('ARC status:', health)
+    prt_i1('Memory throttle count:', throttle)
+    print()
+
+    arc_size = arc_stats['size']
+    arc_target_size = arc_stats['c']
+    arc_max = arc_stats['c_max']
+    arc_min = arc_stats['c_min']
+    mfu_size = arc_stats['mfu_size']
+    mru_size = arc_stats['mru_size']
+    target_size_ratio = '{0}:1'.format(int(arc_max) // int(arc_min))
+
+    prt_2('ARC size (current):',
+          f_perc(arc_size, arc_max), f_bytes(arc_size))
+    prt_i2('Target size (adaptive):',
+           f_perc(arc_target_size, arc_max), f_bytes(arc_target_size))
+    prt_i2('Min size (hard limit):',
+           f_perc(arc_min, arc_max), f_bytes(arc_min))
+    prt_i2('Max size (high water):',
+           target_size_ratio, f_bytes(arc_max))
+    caches_size = int(mfu_size)+int(mru_size)
+    prt_i2('Most Frequently Used (MFU) cache size:',
+           f_perc(mfu_size, caches_size), f_bytes(mfu_size))
+    prt_i2('Most Recently Used (MRU) cache size:',
+           f_perc(mru_size, caches_size), f_bytes(mru_size))
+    print()
+
+    print('ARC hash breakdown:')
+    prt_i1('Elements max:', f_hits(arc_stats['hash_elements_max']))
+    prt_i2('Elements current:',
+           f_perc(arc_stats['hash_elements'], arc_stats['hash_elements_max']),
+           f_hits(arc_stats['hash_elements']))
+    prt_i1('Collisions:', f_hits(arc_stats['hash_collisions']))
+
+    prt_i1('Chain max:', f_hits(arc_stats['hash_chain_max']))
+    prt_i1('Chains:', f_hits(arc_stats['hash_chains']))
+    print()
+
+    print('ARC misc:')
+    prt_i1('Deleted:', f_hits(arc_stats['deleted']))
+    prt_i1('Mutex misses:', f_hits(arc_stats['mutex_miss']))
+    prt_i1('Eviction skips:', f_hits(arc_stats['evict_skip']))
+    print()
+
+
+def section_archits(kstats_dict):
+    """Print information on how the caches are accessed ("arc hits").
+    """
+
+    arc_stats = isolate_section('arcstats', kstats_dict)
+    all_accesses = int(arc_stats['hits'])+int(arc_stats['misses'])
+    actual_hits = int(arc_stats['mfu_hits'])+int(arc_stats['mru_hits'])
+
+    prt_1('ARC total accesses (hits + misses):', f_hits(all_accesses))
+    ta_todo = (('Cache hit ratio:', arc_stats['hits']),
+               ('Cache miss ratio:', arc_stats['misses']),
+               ('Actual hit ratio (MFU + MRU hits):', actual_hits))
+
+    for title, value in ta_todo:
+        prt_i2(title, f_perc(value, all_accesses), f_hits(value))
+
+    dd_total = int(arc_stats['demand_data_hits']) +\
+        int(arc_stats['demand_data_misses'])
+    prt_i2('Data demand efficiency:',
+           f_perc(arc_stats['demand_data_hits'], dd_total),
+           f_hits(dd_total))
+
+    dp_total = int(arc_stats['prefetch_data_hits']) +\
+        int(arc_stats['prefetch_data_misses'])
+    prt_i2('Data prefetch efficiency:',
+           f_perc(arc_stats['prefetch_data_hits'], dp_total),
+           f_hits(dp_total))
+
+    known_hits = int(arc_stats['mfu_hits']) +\
+        int(arc_stats['mru_hits']) +\
+        int(arc_stats['mfu_ghost_hits']) +\
+        int(arc_stats['mru_ghost_hits'])
+
+    anon_hits = int(arc_stats['hits'])-known_hits
+
+    print()
+    print('Cache hits by cache type:')
+    cl_todo = (('Most frequently used (MFU):', arc_stats['mfu_hits']),
+               ('Most recently used (MRU):', arc_stats['mru_hits']),
+               ('Most frequently used (MFU) ghost:',
+                arc_stats['mfu_ghost_hits']),
+               ('Most recently used (MRU) ghost:',
+                arc_stats['mru_ghost_hits']))
+
+    for title, value in cl_todo:
+        prt_i2(title, f_perc(value, arc_stats['hits']), f_hits(value))
+
+    # For some reason, anon_hits can turn negative, which is weird. Until we
+    # have figured out why this happens, we just hide the problem, following
+    # the behavior of the original arc_summary.py
+    if anon_hits >= 0:
+        prt_i2('Anonymously used:',
+               f_perc(anon_hits, arc_stats['hits']), f_hits(anon_hits))
+
+    print()
+    print('Cache hits by data type:')
+    dt_todo = (('Demand data:', arc_stats['demand_data_hits']),
+               ('Demand perfetch data:', arc_stats['prefetch_data_hits']),
+               ('Demand metadata:', arc_stats['demand_metadata_hits']),
+               ('Demand prefetch metadata:',
+                arc_stats['prefetch_metadata_hits']))
+
+    for title, value in dt_todo:
+        prt_i2(title, f_perc(value, arc_stats['hits']), f_hits(value))
+
+    print()
+    print('Cache misses by data type:')
+    dm_todo = (('Demand data:', arc_stats['demand_data_misses']),
+               ('Demand prefetch data:',
+                arc_stats['prefetch_data_misses']),
+               ('Demand metadata:', arc_stats['demand_metadata_misses']),
+               ('Demand prefetch metadata:',
+                arc_stats['prefetch_metadata_misses']))
+
+    for title, value in dm_todo:
+        prt_i2(title, f_perc(value, arc_stats['misses']), f_hits(value))
+
+    print()
+
+
+def section_dmu(kstats_dict):
+    """Collect information on the DMU"""
+
+    zfetch_stats = isolate_section('zfetchstats', kstats_dict)
+
+    zfetch_access_total = int(zfetch_stats['hits'])+int(zfetch_stats['misses'])
+
+    prt_1('DMU prefetch efficiency:', f_hits(zfetch_access_total))
+    prt_i2('Hit ratio:', f_perc(zfetch_stats['hits'], zfetch_access_total),
+           f_hits(zfetch_stats['hits']))
+    prt_i2('Miss ratio:', f_perc(zfetch_stats['misses'], zfetch_access_total),
+           f_hits(zfetch_stats['misses']))
+    print()
+
+
+def section_l2arc(kstats_dict):
+    """Collect information on L2ARC device if present. If not, tell user
+    that we're skipping the section.
+    """
+
+    # The L2ARC statistics live in the same section as the normal ARC stuff
+    arc_stats = isolate_section('arcstats', kstats_dict)
+
+    if arc_stats['l2_size'] == '0':
+        print('L2ARC not detected, skipping section\n')
+        return
+
+    l2_errors = int(arc_stats['l2_writes_error']) +\
+        int(arc_stats['l2_cksum_bad']) +\
+        int(arc_stats['l2_io_error'])
+
+    l2_access_total = int(arc_stats['l2_hits'])+int(arc_stats['l2_misses'])
+    health = 'HEALTHY'
+
+    if l2_errors > 0:
+        health = 'DEGRADED'
+
+    prt_1('L2ARC status:', health)
+
+    l2_todo = (('Low memory aborts:', 'l2_abort_lowmem'),
+               ('Free on write:', 'l2_free_on_write'),
+               ('R/W clashes:', 'l2_rw_clash'),
+               ('Bad checksums:', 'l2_cksum_bad'),
+               ('I/O errors:', 'l2_io_error'))
+
+    for title, value in l2_todo:
+        prt_i1(title, f_hits(arc_stats[value]))
+
+    print()
+    prt_1('L2ARC size (adaptive):', f_bytes(arc_stats['l2_size']))
+    prt_i2('Compressed:', f_perc(arc_stats['l2_asize'], arc_stats['l2_size']),
+           f_bytes(arc_stats['l2_asize']))
+    prt_i2('Header size:',
+           f_perc(arc_stats['l2_hdr_size'], arc_stats['l2_size']),
+           f_bytes(arc_stats['l2_hdr_size']))
+
+    print()
+    prt_1('L2ARC breakdown:', f_hits(l2_access_total))
+    prt_i2('Hit ratio:',
+           f_perc(arc_stats['l2_hits'], l2_access_total),
+           f_bytes(arc_stats['l2_hits']))
+    prt_i2('Miss ratio:',
+           f_perc(arc_stats['l2_misses'], l2_access_total),
+           f_bytes(arc_stats['l2_misses']))
+    prt_i1('Feeds:', f_hits(arc_stats['l2_feeds']))
+
+    print()
+    print('L2ARC writes:')
+
+    if arc_stats['l2_writes_done'] != arc_stats['l2_writes_sent']:
+        prt_i2('Writes sent:', 'FAULTED', f_hits(arc_stats['l2_writes_sent']))
+        prt_i2('Done ratio:',
+               f_perc(arc_stats['l2_writes_done'],
+                      arc_stats['l2_writes_sent']),
+               f_bytes(arc_stats['l2_writes_done']))
+        prt_i2('Error ratio:',
+               f_perc(arc_stats['l2_writes_error'],
+                      arc_stats['l2_writes_sent']),
+               f_bytes(arc_stats['l2_writes_error']))
+    else:
+        prt_i2('Writes sent:', '100 %', f_bytes(arc_stats['l2_writes_sent']))
+
+    print()
+    print('L2ARC evicts:')
+    prt_i1('Lock retries:', f_hits(arc_stats['l2_evict_lock_retry']))
+    prt_i1('Upon reading:', f_hits(arc_stats['l2_evict_reading']))
+    print()
+
+
+def section_spl(*_):
+    """Print the SPL parameters, if requested with alternative format
+    and/or decriptions. This does not use kstats.
+    """
+
+    spls = get_spl_tunables(SPL_PATH)
+    keylist = sorted(spls.keys())
+    print('Solaris Porting Layer (SPL):')
+
+    if ARGS.desc:
+        descriptions = get_descriptions('spl')
+
+    for key in keylist:
+        value = spls[key]
+
+        if ARGS.desc:
+            try:
+                print(INDENT+'#', descriptions[key])
+            except KeyError:
+                print(INDENT+'# (No decription found)')  # paranoid
+
+        print(format_raw_line(key, value))
+
+    print()
+
+
+def section_tunables(*_):
+    """Print the tunables, if requested with alternative format and/or
+    decriptions. This does not use kstasts.
+    """
+
+    tunables = get_spl_tunables(TUNABLES_PATH)
+    keylist = sorted(tunables.keys())
+    print('Tunables:')
+
+    if ARGS.desc:
+        descriptions = get_descriptions('zfs')
+
+    for key in keylist:
+        value = tunables[key]
+
+        if ARGS.desc:
+            try:
+                print(INDENT+'#', descriptions[key])
+            except KeyError:
+                print(INDENT+'# (No decription found)')  # paranoid
+
+        print(format_raw_line(key, value))
+
+    print()
+
+
+def section_vdev(kstats_dict):
+    """Collect information on VDEV caches"""
+
+    # Currently [Nov 2017] the VDEV cache is disabled, because it is actually
+    # harmful. When this is the case, we just skip the whole entry. See
+    # https://github.com/zfsonlinux/zfs/blob/master/module/zfs/vdev_cache.c
+    # for details
+    tunables = get_spl_tunables(TUNABLES_PATH)
+
+    if tunables['zfs_vdev_cache_size'] == '0':
+        print('VDEV cache disabled, skipping section\n')
+        return
+
+    vdev_stats = isolate_section('vdev_cache_stats', kstats_dict)
+
+    vdev_cache_total = int(vdev_stats['hits']) +\
+        int(vdev_stats['misses']) +\
+        int(vdev_stats['delegations'])
+
+    prt_1('VDEV cache summary:', f_hits(vdev_cache_total))
+    prt_i2('Hit ratio:', f_perc(vdev_stats['hits'], vdev_cache_total),
+           f_hits(vdev_stats['hits']))
+    prt_i2('Miss ratio:', f_perc(vdev_stats['misses'], vdev_cache_total),
+           f_hits(vdev_stats['misses']))
+    prt_i2('Delegations:', f_perc(vdev_stats['delegations'], vdev_cache_total),
+           f_hits(vdev_stats['delegations']))
+    print()
+
+
+def section_zil(kstats_dict):
+    """Collect information on the ZFS Intent Log. Some of the information
+    taken from https://github.com/zfsonlinux/zfs/blob/master/include/sys/zil.h
+    """
+
+    zil_stats = isolate_section('zil', kstats_dict)
+
+    prt_1('ZIL committed transactions:',
+          f_hits(zil_stats['zil_itx_count']))
+    prt_i1('Commit requests:', f_hits(zil_stats['zil_commit_count']))
+    prt_i1('Flushes to stable storage:',
+           f_hits(zil_stats['zil_commit_writer_count']))
+    prt_i2('Transactions to SLOG storage pool:',
+           f_bytes(zil_stats['zil_itx_metaslab_slog_bytes']),
+           f_hits(zil_stats['zil_itx_metaslab_slog_count']))
+    prt_i2('Transactions to non-SLOG storage pool:',
+           f_bytes(zil_stats['zil_itx_metaslab_normal_bytes']),
+           f_hits(zil_stats['zil_itx_metaslab_normal_count']))
+    print()
+
+
+section_calls = {'arc': section_arc,
+                 'archits': section_archits,
+                 'dmu': section_dmu,
+                 'l2arc': section_l2arc,
+                 'spl': section_spl,
+                 'tunables': section_tunables,
+                 'vdev': section_vdev,
+                 'zil': section_zil}
+
+
+def main():
+    """Run program. The options to draw a graph and to print all data raw are
+    treated separately because they come with their own call.
+    """
+
+    kstats = get_kstats()
+
+    if ARGS.graph:
+        draw_graph(kstats)
+        sys.exit(0)
+
+    print_header()
+
+    if ARGS.raw:
+        print_raw(kstats)
+
+    elif ARGS.section:
+
+        try:
+            section_calls[ARGS.section](kstats)
+        except KeyError:
+            print('Error: Section "{0}" unknown'.format(ARGS.section))
+            sys.exit(1)
+
+    elif ARGS.page:
+        print('WARNING: Pages are deprecated, please use "--section"\n')
+
+        pages_to_calls = {1: 'arc',
+                          2: 'archits',
+                          3: 'l2arc',
+                          4: 'dmu',
+                          5: 'vdev',
+                          6: 'tunables'}
+
+        try:
+            call = pages_to_calls[ARGS.page]
+        except KeyError:
+            print('Error: Page "{0}" not supported'.format(ARGS.page))
+            sys.exit(1)
+        else:
+            section_calls[call](kstats)
+
+    else:
+        # If no parameters were given, we print all sections. We might want to
+        # change the sequence by hand
+        calls = sorted(section_calls.keys())
+
+        for section in calls:
+            section_calls[section](kstats)
+
+    sys.exit(0)
+
+
+if __name__ == '__main__':
+    main()

tests/runfiles/linux.run

@@ -436,7 +436,8 @@ tests = ['zdb_001_neg', 'zfs_001_neg', 'zfs_allow_001_neg',
     'zpool_offline_001_neg', 'zpool_online_001_neg', 'zpool_remove_001_neg',
     'zpool_replace_001_neg', 'zpool_scrub_001_neg', 'zpool_set_001_neg',
     'zpool_status_001_neg', 'zpool_upgrade_001_neg', 'arcstat_001_pos',
-    'arc_summary_001_pos', 'arc_summary_002_neg', 'dbufstat_001_pos']
+    'arc_summary_001_pos', 'arc_summary_002_neg', 
+    'arc_summary3_001_pos', 'dbufstat_001_pos']
 user =
 tags = ['functional', 'cli_user', 'misc']
 

tests/zfs-tests/include/commands.cfg

@@ -91,6 +91,7 @@ export SYSTEM_FILES='arp
     ps
     pwd
     python
+    python3
     quotaon
     readlink
     rm
@@ -143,6 +144,7 @@ export ZFS_FILES='zdb
     ztest
     raidz_test
     arc_summary.py
+    arc_summary3.py
     arcstat.py
     dbufstat.py
     zed

tests/zfs-tests/tests/functional/cli_user/misc/Makefile.am

@@ -47,4 +47,5 @@ dist_pkgdata_SCRIPTS = \
 	arcstat_001_pos.ksh \
 	arc_summary_001_pos.ksh \
 	arc_summary_002_neg.ksh \
+	arc_summary3_001_pos.ksh \
 	dbufstat_001_pos.ksh

tests/zfs-tests/tests/functional/cli_user/misc/arc_summary3_001_pos.ksh

@@ -0,0 +1,56 @@
+#! /bin/ksh -p
+#
+# CDDL HEADER START
+#
+# The contents of this file are subject to the terms of the
+# Common Development and Distribution License (the "License").
+# You may not use this file except in compliance with the License.
+#
+# You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+# or http://www.opensolaris.org/os/licensing.
+# See the License for the specific language governing permissions
+# and limitations under the License.
+#
+# When distributing Covered Code, include this CDDL HEADER in each
+# file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+# If applicable, add the following below this CDDL HEADER, with the
+# fields enclosed by brackets "[]" replaced with your own identifying
+# information: Portions Copyright [yyyy] [name of copyright owner]
+#
+# CDDL HEADER END
+#
+
+#
+# Copyright (c) 2015 by Lawrence Livermore National Security, LLC.
+# All rights reserved.
+#
+
+. $STF_SUITE/include/libtest.shlib
+
+# Keep the following test until Python 3 is installed on all test systems,
+# then remove
+python3 -V 2>&1 > /dev/null
+if (( $? )); then
+        log_unsupported "Python3 is not installed"
+fi
+
+
+# Some systems have Python 3 installed, but only older versions that don't
+# have the subprocess.run() functionality. We catch these with a separate
+# test. Remove this when all systems have reached 3.5 or greater
+VERSIONPYTEST=$(python3 -V)
+if [[ ${VERSIONPYTEST:9:1} -lt 5 ]]; then
+        log_unsupported "Python3 must be version 3.5 or greater"
+fi
+
+
+set -A args  "" "-a" "-d" "-p 1" "-g" "-s arc" "-r"
+log_assert "arc_summary3.py generates output and doesn't return an error code"
+
+typeset -i i=0
+while [[ $i -lt ${#args[*]} ]]; do
+        log_must eval "arc_summary3.py ${args[i]} > /dev/null"
+        ((i = i + 1))
+done
+
+log_pass "arc_summary3.py generates output and doesn't return an error code"