mirror_zfs/cmd/arcstat.in

675 lines
23 KiB
Plaintext
Raw Normal View History

#!/usr/bin/env @PYTHON_SHEBANG@
#
# Print out ZFS ARC Statistics exported via kstat(1)
# For a definition of fields, or usage, use arcstat -v
#
# This script was originally a fork of the original arcstat.pl (0.1)
# by Neelakanth Nadgir, originally published on his Sun blog on
# 09/18/2007
# http://blogs.sun.com/realneel/entry/zfs_arc_statistics
#
# A new version aimed to improve upon the original by adding features
# and fixing bugs as needed. This version was maintained by Mike
# Harsch and was hosted in a public open source repository:
# http://github.com/mharsch/arcstat
#
# but has since moved to the illumos-gate repository.
#
# This Python port was written by John Hixson for FreeNAS, introduced
# in commit e2c29f:
# https://github.com/freenas/freenas
#
# and has been improved by many people since.
#
# CDDL HEADER START
#
# The contents of this file are subject to the terms of the
# Common Development and Distribution License, Version 1.0 only
# (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 https://opensource.org/licenses/CDDL-1.0.
# 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
#
#
# Fields have a fixed width. Every interval, we fill the "v"
# hash with its corresponding value (v[field]=value) using calculate().
# @hdr is the array of fields that needs to be printed, so we
# just iterate over this array and print the values using our pretty printer.
#
# This script must remain compatible with Python 3.6+.
pyzfs: python3 support (build system) Almost all of the Python code in the respository has been updated to be compatibile with Python 2.6, Python 3.4, or newer. The only exceptions are arc_summery3.py which requires Python 3, and pyzfs which requires at least Python 2.7. This allows us to maintain a single version of the code and support most default versions of python. This change does the following: * Sets the default shebang for all Python scripts to python3. If only Python 2 is available, then at install time scripts which are compatible with Python 2 will have their shebangs replaced with /usr/bin/python. This is done for compatibility until Python 2 goes end of life. Since only the installed versions are changed this means Python 3 must be installed on the system for test-runner when testing in-tree. * Added --with-python=<2|3|3.4,etc> configure option which sets the PYTHON environment variable to target a specific python version. By default the newest installed version of Python will be used or the preferred distribution version when creating pacakges. * Fixed --enable-pyzfs configure checks so they are run when --enable-pyzfs=check and --enable-pyzfs=yes. * Enabled pyzfs for Python 3.4 and newer, which is now supported. * Renamed pyzfs package to python<VERSION>-pyzfs and updated to install in the appropriate site location. For example, when building with --with-python=3.4 a python34-pyzfs will be created which installs in /usr/lib/python3.4/site-packages/. * Renamed the following python scripts according to the Fedora guidance for packaging utilities in /bin - dbufstat.py -> dbufstat - arcstat.py -> arcstat - arc_summary.py -> arc_summary - arc_summary3.py -> arc_summary3 * Updated python-cffi package name. On CentOS 6, CentOS 7, and Amazon Linux it's called python-cffi, not python2-cffi. For Python3 it's called python3-cffi or python3x-cffi. * Install one version of arc_summary. Depending on the version of Python available install either arc_summary2 or arc_summary3 as arc_summary. The user output is only slightly different. Reviewed-by: John Ramsden <johnramsden@riseup.net> Reviewed-by: Neal Gompa <ngompa@datto.com> Reviewed-by: loli10K <ezomori.nozomu@gmail.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #8096
2018-10-31 19:22:59 +03:00
#
import sys
import time
import getopt
import re
import copy
from signal import signal, SIGINT, SIGWINCH, SIG_DFL
cols = {
# HDR: [Size, Scale, Description]
"time": [8, -1, "Time"],
"hits": [4, 1000, "ARC hits per second"],
"iohs": [4, 1000, "ARC I/O hits per second"],
"miss": [4, 1000, "ARC misses per second"],
"read": [4, 1000, "Total ARC accesses per second"],
"hit%": [4, 100, "ARC hit percentage"],
"ioh%": [4, 100, "ARC I/O hit percentage"],
"miss%": [5, 100, "ARC miss percentage"],
"dhit": [4, 1000, "Demand hits per second"],
"dioh": [4, 1000, "Demand I/O hits per second"],
"dmis": [4, 1000, "Demand misses per second"],
"dh%": [3, 100, "Demand hit percentage"],
"di%": [3, 100, "Demand I/O hit percentage"],
"dm%": [3, 100, "Demand miss percentage"],
"ddhit": [5, 1000, "Demand data hits per second"],
"ddioh": [5, 1000, "Demand data I/O hits per second"],
"ddmis": [5, 1000, "Demand data misses per second"],
"ddh%": [4, 100, "Demand data hit percentage"],
"ddi%": [4, 100, "Demand data I/O hit percentage"],
"ddm%": [4, 100, "Demand data miss percentage"],
"dmhit": [5, 1000, "Demand metadata hits per second"],
"dmioh": [5, 1000, "Demand metadata I/O hits per second"],
"dmmis": [5, 1000, "Demand metadata misses per second"],
"dmh%": [4, 100, "Demand metadata hit percentage"],
"dmi%": [4, 100, "Demand metadata I/O hit percentage"],
"dmm%": [4, 100, "Demand metadata miss percentage"],
"phit": [4, 1000, "Prefetch hits per second"],
"pioh": [4, 1000, "Prefetch I/O hits per second"],
"pmis": [4, 1000, "Prefetch misses per second"],
"ph%": [3, 100, "Prefetch hits percentage"],
"pi%": [3, 100, "Prefetch I/O hits percentage"],
"pm%": [3, 100, "Prefetch miss percentage"],
"pdhit": [5, 1000, "Prefetch data hits per second"],
"pdioh": [5, 1000, "Prefetch data I/O hits per second"],
"pdmis": [5, 1000, "Prefetch data misses per second"],
"pdh%": [4, 100, "Prefetch data hits percentage"],
"pdi%": [4, 100, "Prefetch data I/O hits percentage"],
"pdm%": [4, 100, "Prefetch data miss percentage"],
"pmhit": [5, 1000, "Prefetch metadata hits per second"],
"pmioh": [5, 1000, "Prefetch metadata I/O hits per second"],
"pmmis": [5, 1000, "Prefetch metadata misses per second"],
"pmh%": [4, 100, "Prefetch metadata hits percentage"],
"pmi%": [4, 100, "Prefetch metadata I/O hits percentage"],
"pmm%": [4, 100, "Prefetch metadata miss percentage"],
"mhit": [4, 1000, "Metadata hits per second"],
"mioh": [4, 1000, "Metadata I/O hits per second"],
"mmis": [4, 1000, "Metadata misses per second"],
"mread": [5, 1000, "Metadata accesses per second"],
"mh%": [3, 100, "Metadata hit percentage"],
"mi%": [3, 100, "Metadata I/O hit percentage"],
"mm%": [3, 100, "Metadata miss percentage"],
"arcsz": [5, 1024, "ARC size"],
"size": [5, 1024, "ARC size"],
"c": [5, 1024, "ARC target size"],
"mfu": [4, 1000, "MFU list hits per second"],
"mru": [4, 1000, "MRU list hits per second"],
"mfug": [4, 1000, "MFU ghost list hits per second"],
"mrug": [4, 1000, "MRU ghost list hits per second"],
"unc": [4, 1000, "Uncached list hits per second"],
"eskip": [5, 1000, "evict_skip per second"],
Add L2ARC arcstats for MFU/MRU buffers and buffer content type Currently the ARC state (MFU/MRU) of cached L2ARC buffer and their content type is unknown. Knowing this information may prove beneficial in adjusting the L2ARC caching policy. This commit adds L2ARC arcstats that display the aligned size (in bytes) of L2ARC buffers according to their content type (data/metadata) and according to their ARC state (MRU/MFU or prefetch). It also expands the existing evict_l2_eligible arcstat to differentiate between MFU and MRU buffers. L2ARC caches buffers from the MRU and MFU lists of ARC. Upon caching a buffer, its ARC state (MRU/MFU) is stored in the L2 header (b_arcs_state). The l2_m{f,r}u_asize arcstats reflect the aligned size (in bytes) of L2ARC buffers according to their ARC state (based on b_arcs_state). We also account for the case where an L2ARC and ARC cached MRU or MRU_ghost buffer transitions to MFU. The l2_prefetch_asize reflects the alinged size (in bytes) of L2ARC buffers that were cached while they had the prefetch flag set in ARC. This is dynamically updated as the prefetch flag of L2ARC buffers changes. When buffers are evicted from ARC, if they are determined to be L2ARC eligible then their logical size is recorded in evict_l2_eligible_m{r,f}u arcstats according to their ARC state upon eviction. Persistent L2ARC: When committing an L2ARC buffer to a log block (L2ARC metadata) its b_arcs_state and prefetch flag is also stored. If the buffer changes its arcstate or prefetch flag this is reflected in the above arcstats. However, the L2ARC metadata cannot currently be updated to reflect this change. Example: L2ARC caches an MRU buffer. L2ARC metadata and arcstats count this as an MRU buffer. The buffer transitions to MFU. The arcstats are updated to reflect this. Upon pool re-import or on/offlining the L2ARC device the arcstats are cleared and the buffer will now be counted as an MRU buffer, as the L2ARC metadata were not updated. Bug fix: - If l2arc_noprefetch is set, arc_read_done clears the L2CACHE flag of an ARC buffer. However, prefetches may be issued in a way that arc_read_done() is bypassed. Instead, move the related code in l2arc_write_eligible() to account for those cases too. Also add a test and update manpages for l2arc_mfuonly module parameter, and update the manpages and code comments for l2arc_noprefetch. Move persist_l2arc tests to l2arc. Reviewed-by: Ryan Moeller <freqlabs@FreeBSD.org> Reviewed-by: Richard Elling <Richard.Elling@RichardElling.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: George Amanakis <gamanakis@gmail.com> Closes #10743
2020-09-14 20:10:44 +03:00
"el2skip": [7, 1000, "evict skip, due to l2 writes, per second"],
"el2cach": [7, 1024, "Size of L2 cached evictions per second"],
"el2el": [5, 1024, "Size of L2 eligible evictions per second"],
"el2mfu": [6, 1024, "Size of L2 eligible MFU evictions per second"],
"el2mru": [6, 1024, "Size of L2 eligible MRU evictions per second"],
"el2inel": [7, 1024, "Size of L2 ineligible evictions per second"],
"mtxmis": [6, 1000, "mutex_miss per second"],
"dread": [5, 1000, "Demand accesses per second"],
"ddread": [6, 1000, "Demand data accesses per second"],
"dmread": [6, 1000, "Demand metadata accesses per second"],
"pread": [5, 1000, "Prefetch accesses per second"],
"pdread": [6, 1000, "Prefetch data accesses per second"],
"pmread": [6, 1000, "Prefetch metadata accesses per second"],
"l2hits": [6, 1000, "L2ARC hits per second"],
"l2miss": [6, 1000, "L2ARC misses per second"],
"l2read": [6, 1000, "Total L2ARC accesses per second"],
"l2hit%": [6, 100, "L2ARC access hit percentage"],
"l2miss%": [7, 100, "L2ARC access miss percentage"],
Add L2ARC arcstats for MFU/MRU buffers and buffer content type Currently the ARC state (MFU/MRU) of cached L2ARC buffer and their content type is unknown. Knowing this information may prove beneficial in adjusting the L2ARC caching policy. This commit adds L2ARC arcstats that display the aligned size (in bytes) of L2ARC buffers according to their content type (data/metadata) and according to their ARC state (MRU/MFU or prefetch). It also expands the existing evict_l2_eligible arcstat to differentiate between MFU and MRU buffers. L2ARC caches buffers from the MRU and MFU lists of ARC. Upon caching a buffer, its ARC state (MRU/MFU) is stored in the L2 header (b_arcs_state). The l2_m{f,r}u_asize arcstats reflect the aligned size (in bytes) of L2ARC buffers according to their ARC state (based on b_arcs_state). We also account for the case where an L2ARC and ARC cached MRU or MRU_ghost buffer transitions to MFU. The l2_prefetch_asize reflects the alinged size (in bytes) of L2ARC buffers that were cached while they had the prefetch flag set in ARC. This is dynamically updated as the prefetch flag of L2ARC buffers changes. When buffers are evicted from ARC, if they are determined to be L2ARC eligible then their logical size is recorded in evict_l2_eligible_m{r,f}u arcstats according to their ARC state upon eviction. Persistent L2ARC: When committing an L2ARC buffer to a log block (L2ARC metadata) its b_arcs_state and prefetch flag is also stored. If the buffer changes its arcstate or prefetch flag this is reflected in the above arcstats. However, the L2ARC metadata cannot currently be updated to reflect this change. Example: L2ARC caches an MRU buffer. L2ARC metadata and arcstats count this as an MRU buffer. The buffer transitions to MFU. The arcstats are updated to reflect this. Upon pool re-import or on/offlining the L2ARC device the arcstats are cleared and the buffer will now be counted as an MRU buffer, as the L2ARC metadata were not updated. Bug fix: - If l2arc_noprefetch is set, arc_read_done clears the L2CACHE flag of an ARC buffer. However, prefetches may be issued in a way that arc_read_done() is bypassed. Instead, move the related code in l2arc_write_eligible() to account for those cases too. Also add a test and update manpages for l2arc_mfuonly module parameter, and update the manpages and code comments for l2arc_noprefetch. Move persist_l2arc tests to l2arc. Reviewed-by: Ryan Moeller <freqlabs@FreeBSD.org> Reviewed-by: Richard Elling <Richard.Elling@RichardElling.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: George Amanakis <gamanakis@gmail.com> Closes #10743
2020-09-14 20:10:44 +03:00
"l2pref": [6, 1024, "L2ARC prefetch allocated size"],
"l2mfu": [5, 1024, "L2ARC MFU allocated size"],
"l2mru": [5, 1024, "L2ARC MRU allocated size"],
"l2data": [6, 1024, "L2ARC data allocated size"],
"l2meta": [6, 1024, "L2ARC metadata allocated size"],
"l2pref%": [7, 100, "L2ARC prefetch percentage"],
"l2mfu%": [6, 100, "L2ARC MFU percentage"],
"l2mru%": [6, 100, "L2ARC MRU percentage"],
"l2data%": [7, 100, "L2ARC data percentage"],
"l2meta%": [7, 100, "L2ARC metadata percentage"],
"l2asize": [7, 1024, "Actual (compressed) size of the L2ARC"],
"l2size": [6, 1024, "Size of the L2ARC"],
"l2bytes": [7, 1024, "Bytes read per second from the L2ARC"],
"grow": [4, 1000, "ARC grow disabled"],
"need": [5, 1024, "ARC reclaim need"],
"free": [5, 1024, "ARC free memory"],
"avail": [5, 1024, "ARC available memory"],
Include scatter_chunk_waste in arc_size The ARC caches data in scatter ABD's, which are collections of pages, which are typically 4K. Therefore, the space used to cache each block is rounded up to a multiple of 4K. The ABD subsystem tracks this wasted memory in the `scatter_chunk_waste` kstat. However, the ARC's `size` is not aware of the memory used by this round-up, it only accounts for the size that it requested from the ABD subsystem. Therefore, the ARC is effectively using more memory than it is aware of, due to the `scatter_chunk_waste`. This impacts observability, e.g. `arcstat` will show that the ARC is using less memory than it effectively is. It also impacts how the ARC responds to memory pressure. As the amount of `scatter_chunk_waste` changes, it appears to the ARC as memory pressure, so it needs to resize `arc_c`. If the sector size (`1<<ashift`) is the same as the page size (or larger), there won't be any waste. If the (compressed) block size is relatively large compared to the page size, the amount of `scatter_chunk_waste` will be small, so the problematic effects are minimal. However, if using 512B sectors (`ashift=9`), and the (compressed) block size is small (e.g. `compression=on` with the default `volblocksize=8k` or a decreased `recordsize`), the amount of `scatter_chunk_waste` can be very large. On a production system, with `arc_size` at a constant 50% of memory, `scatter_chunk_waste` has been been observed to be 10-30% of memory. This commit adds `scatter_chunk_waste` to `arc_size`, and adds a new `waste` field to `arcstat`. As a result, the ARC's memory usage is more observable, and `arc_c` does not need to be adjusted as frequently. Reviewed-by: Pavel Zakharov <pavel.zakharov@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: George Wilson <gwilson@delphix.com> Reviewed-by: Ryan Moeller <ryan@iXsystems.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #10701
2020-08-18 06:04:04 +03:00
"waste": [5, 1024, "Wasted memory due to round up to pagesize"],
"ztotal": [6, 1000, "zfetch total prefetcher calls per second"],
"zhits": [5, 1000, "zfetch stream hits per second"],
"zahead": [6, 1000, "zfetch hits ahead of streams per second"],
"zpast": [5, 1000, "zfetch hits behind streams per second"],
"zmisses": [7, 1000, "zfetch stream misses per second"],
"zmax": [4, 1000, "zfetch limit reached per second"],
"zfuture": [7, 1000, "zfetch stream future per second"],
"zstride": [7, 1000, "zfetch stream strides per second"],
"zissued": [7, 1000, "zfetch prefetches issued per second"],
"zactive": [7, 1000, "zfetch prefetches active per second"],
}
v = {}
hdr = ["time", "read", "ddread", "ddh%", "dmread", "dmh%", "pread", "ph%",
"size", "c", "avail"]
xhdr = ["time", "mfu", "mru", "mfug", "mrug", "unc", "eskip", "mtxmis",
"dread", "pread", "read"]
zhdr = ["time", "ztotal", "zhits", "zahead", "zpast", "zmisses", "zmax",
"zfuture", "zstride", "zissued", "zactive"]
sint = 1 # Default interval is 1 second
count = 1 # Default count is 1
hdr_intr = 20 # Print header every 20 lines of output
opfile = None
sep = " " # Default separator is 2 spaces
l2exist = False
cmd = ("Usage: arcstat [-havxp] [-f fields] [-o file] [-s string] [interval "
"[count]]\n")
cur = {}
d = {}
out = None
kstat = None
pretty_print = True
if sys.platform.startswith('freebsd'):
# Requires py-sysctl on FreeBSD
import sysctl
def kstat_update():
global kstat
k = [ctl for ctl in sysctl.filter('kstat.zfs.misc.arcstats')
if ctl.type != sysctl.CTLTYPE_NODE]
if not k:
sys.exit(1)
kstat = {}
for s in k:
if not s:
continue
name, value = s.name, s.value
# Trims 'kstat.zfs.misc.arcstats' from the name
kstat[name[24:]] = int(value)
elif sys.platform.startswith('linux'):
def kstat_update():
global kstat
k1 = [line.strip() for line in open('/proc/spl/kstat/zfs/arcstats')]
k2 = ["zfetch_" + line.strip() for line in
open('/proc/spl/kstat/zfs/zfetchstats')]
if k1 is None or k2 is None:
sys.exit(1)
del k1[0:2]
del k2[0:2]
k = k1 + k2
kstat = {}
for s in k:
if not s:
continue
name, unused, value = s.split()
kstat[name] = int(value)
def detailed_usage():
sys.stderr.write("%s\n" % cmd)
sys.stderr.write("Field definitions are as follows:\n")
for key in cols:
sys.stderr.write("%11s : %s\n" % (key, cols[key][2]))
sys.stderr.write("\n")
sys.exit(0)
def usage():
sys.stderr.write("%s\n" % cmd)
sys.stderr.write("\t -h : Print this help message\n")
sys.stderr.write("\t -a : Print all possible stats\n")
sys.stderr.write("\t -v : List all possible field headers and definitions"
"\n")
sys.stderr.write("\t -x : Print extended stats\n")
sys.stderr.write("\t -z : Print zfetch stats\n")
sys.stderr.write("\t -f : Specify specific fields to print (see -v)\n")
sys.stderr.write("\t -o : Redirect output to the specified file\n")
sys.stderr.write("\t -s : Override default field separator with custom "
"character or string\n")
sys.stderr.write("\t -p : Disable auto-scaling of numerical fields\n")
sys.stderr.write("\nExamples:\n")
pyzfs: python3 support (build system) Almost all of the Python code in the respository has been updated to be compatibile with Python 2.6, Python 3.4, or newer. The only exceptions are arc_summery3.py which requires Python 3, and pyzfs which requires at least Python 2.7. This allows us to maintain a single version of the code and support most default versions of python. This change does the following: * Sets the default shebang for all Python scripts to python3. If only Python 2 is available, then at install time scripts which are compatible with Python 2 will have their shebangs replaced with /usr/bin/python. This is done for compatibility until Python 2 goes end of life. Since only the installed versions are changed this means Python 3 must be installed on the system for test-runner when testing in-tree. * Added --with-python=<2|3|3.4,etc> configure option which sets the PYTHON environment variable to target a specific python version. By default the newest installed version of Python will be used or the preferred distribution version when creating pacakges. * Fixed --enable-pyzfs configure checks so they are run when --enable-pyzfs=check and --enable-pyzfs=yes. * Enabled pyzfs for Python 3.4 and newer, which is now supported. * Renamed pyzfs package to python<VERSION>-pyzfs and updated to install in the appropriate site location. For example, when building with --with-python=3.4 a python34-pyzfs will be created which installs in /usr/lib/python3.4/site-packages/. * Renamed the following python scripts according to the Fedora guidance for packaging utilities in /bin - dbufstat.py -> dbufstat - arcstat.py -> arcstat - arc_summary.py -> arc_summary - arc_summary3.py -> arc_summary3 * Updated python-cffi package name. On CentOS 6, CentOS 7, and Amazon Linux it's called python-cffi, not python2-cffi. For Python3 it's called python3-cffi or python3x-cffi. * Install one version of arc_summary. Depending on the version of Python available install either arc_summary2 or arc_summary3 as arc_summary. The user output is only slightly different. Reviewed-by: John Ramsden <johnramsden@riseup.net> Reviewed-by: Neal Gompa <ngompa@datto.com> Reviewed-by: loli10K <ezomori.nozomu@gmail.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #8096
2018-10-31 19:22:59 +03:00
sys.stderr.write("\tarcstat -o /tmp/a.log 2 10\n")
sys.stderr.write("\tarcstat -s \",\" -o /tmp/a.log 2 10\n")
sys.stderr.write("\tarcstat -v\n")
sys.stderr.write("\tarcstat -f time,hit%,dh%,ph%,mh% 1\n")
sys.stderr.write("\n")
sys.exit(1)
def snap_stats():
global cur
global kstat
prev = copy.deepcopy(cur)
kstat_update()
cur = kstat
for key in cur:
if re.match(key, "class"):
continue
if key in prev:
d[key] = cur[key] - prev[key]
else:
d[key] = cur[key]
def prettynum(sz, scale, num=0):
suffix = [' ', 'K', 'M', 'G', 'T', 'P', 'E', 'Z']
index = 0
save = 0
# Special case for date field
if scale == -1:
return "%s" % num
# Rounding error, return 0
elif 0 < num < 1:
num = 0
while abs(num) > scale and index < 5:
save = num
num = num / scale
index += 1
if index == 0:
return "%*d" % (sz, num)
if abs(save / scale) < 10:
return "%*.1f%s" % (sz - 1, num, suffix[index])
else:
return "%*d%s" % (sz - 1, num, suffix[index])
def print_values():
global hdr
global sep
global v
global pretty_print
if pretty_print:
fmt = lambda col: prettynum(cols[col][0], cols[col][1], v[col])
else:
fmt = lambda col: str(v[col])
sys.stdout.write(sep.join(fmt(col) for col in hdr))
sys.stdout.write("\n")
sys.stdout.flush()
def print_header():
global hdr
global sep
global pretty_print
if pretty_print:
fmt = lambda col: "%*s" % (cols[col][0], col)
else:
fmt = lambda col: col
sys.stdout.write(sep.join(fmt(col) for col in hdr))
sys.stdout.write("\n")
def get_terminal_lines():
try:
import fcntl
import termios
import struct
data = fcntl.ioctl(sys.stdout.fileno(), termios.TIOCGWINSZ, '1234')
sz = struct.unpack('hh', data)
return sz[0]
except Exception:
pass
def update_hdr_intr():
global hdr_intr
lines = get_terminal_lines()
if lines and lines > 3:
hdr_intr = lines - 3
def resize_handler(signum, frame):
update_hdr_intr()
def init():
global sint
global count
global hdr
global xhdr
global zhdr
global opfile
global sep
global out
global l2exist
global pretty_print
desired_cols = None
aflag = False
xflag = False
hflag = False
vflag = False
zflag = False
i = 1
try:
opts, args = getopt.getopt(
sys.argv[1:],
"axzo:hvs:f:p",
[
"all",
"extended",
"zfetch",
"outfile",
"help",
"verbose",
"separator",
"columns",
"parsable"
]
)
except getopt.error as msg:
sys.stderr.write("Error: %s\n" % str(msg))
usage()
opts = None
for opt, arg in opts:
if opt in ('-a', '--all'):
aflag = True
if opt in ('-x', '--extended'):
xflag = True
if opt in ('-o', '--outfile'):
opfile = arg
i += 1
if opt in ('-h', '--help'):
hflag = True
if opt in ('-v', '--verbose'):
vflag = True
if opt in ('-s', '--separator'):
sep = arg
i += 1
if opt in ('-f', '--columns'):
desired_cols = arg
i += 1
if opt in ('-p', '--parsable'):
pretty_print = False
if opt in ('-z', '--zfetch'):
zflag = True
i += 1
argv = sys.argv[i:]
sint = int(argv[0]) if argv else sint
count = int(argv[1]) if len(argv) > 1 else (0 if len(argv) > 0 else 1)
if hflag or (xflag and zflag) or ((zflag or xflag) and desired_cols):
usage()
if vflag:
detailed_usage()
if xflag:
hdr = xhdr
if zflag:
hdr = zhdr
update_hdr_intr()
# check if L2ARC exists
snap_stats()
l2_size = cur.get("l2_size")
if l2_size:
l2exist = True
if desired_cols:
hdr = desired_cols.split(",")
invalid = []
incompat = []
for ele in hdr:
if ele not in cols:
invalid.append(ele)
elif not l2exist and ele.startswith("l2"):
sys.stdout.write("No L2ARC Here\n%s\n" % ele)
incompat.append(ele)
if len(invalid) > 0:
sys.stderr.write("Invalid column definition! -- %s\n" % invalid)
usage()
if len(incompat) > 0:
sys.stderr.write("Incompatible field specified! -- %s\n" %
incompat)
usage()
if aflag:
if l2exist:
hdr = cols.keys()
else:
hdr = [col for col in cols.keys() if not col.startswith("l2")]
if opfile:
try:
out = open(opfile, "w")
sys.stdout = out
except IOError:
sys.stderr.write("Cannot open %s for writing\n" % opfile)
sys.exit(1)
def calculate():
global d
global v
global l2exist
v = dict()
v["time"] = time.strftime("%H:%M:%S", time.localtime())
v["hits"] = d["hits"] // sint
v["iohs"] = d["iohits"] // sint
v["miss"] = d["misses"] // sint
v["read"] = v["hits"] + v["iohs"] + v["miss"]
v["hit%"] = 100 * v["hits"] // v["read"] if v["read"] > 0 else 0
v["ioh%"] = 100 * v["iohs"] // v["read"] if v["read"] > 0 else 0
v["miss%"] = 100 - v["hit%"] - v["ioh%"] if v["read"] > 0 else 0
v["dhit"] = (d["demand_data_hits"] + d["demand_metadata_hits"]) // sint
v["dioh"] = (d["demand_data_iohits"] + d["demand_metadata_iohits"]) // sint
v["dmis"] = (d["demand_data_misses"] + d["demand_metadata_misses"]) // sint
v["dread"] = v["dhit"] + v["dioh"] + v["dmis"]
v["dh%"] = 100 * v["dhit"] // v["dread"] if v["dread"] > 0 else 0
v["di%"] = 100 * v["dioh"] // v["dread"] if v["dread"] > 0 else 0
v["dm%"] = 100 - v["dh%"] - v["di%"] if v["dread"] > 0 else 0
v["ddhit"] = d["demand_data_hits"] // sint
v["ddioh"] = d["demand_data_iohits"] // sint
v["ddmis"] = d["demand_data_misses"] // sint
v["ddread"] = v["ddhit"] + v["ddioh"] + v["ddmis"]
v["ddh%"] = 100 * v["ddhit"] // v["ddread"] if v["ddread"] > 0 else 0
v["ddi%"] = 100 * v["ddioh"] // v["ddread"] if v["ddread"] > 0 else 0
v["ddm%"] = 100 - v["ddh%"] - v["ddi%"] if v["ddread"] > 0 else 0
v["dmhit"] = d["demand_metadata_hits"] // sint
v["dmioh"] = d["demand_metadata_iohits"] // sint
v["dmmis"] = d["demand_metadata_misses"] // sint
v["dmread"] = v["dmhit"] + v["dmioh"] + v["dmmis"]
v["dmh%"] = 100 * v["dmhit"] // v["dmread"] if v["dmread"] > 0 else 0
v["dmi%"] = 100 * v["dmioh"] // v["dmread"] if v["dmread"] > 0 else 0
v["dmm%"] = 100 - v["dmh%"] - v["dmi%"] if v["dmread"] > 0 else 0
v["phit"] = (d["prefetch_data_hits"] + d["prefetch_metadata_hits"]) // sint
v["pioh"] = (d["prefetch_data_iohits"] +
d["prefetch_metadata_iohits"]) // sint
v["pmis"] = (d["prefetch_data_misses"] +
d["prefetch_metadata_misses"]) // sint
v["pread"] = v["phit"] + v["pioh"] + v["pmis"]
v["ph%"] = 100 * v["phit"] // v["pread"] if v["pread"] > 0 else 0
v["pi%"] = 100 * v["pioh"] // v["pread"] if v["pread"] > 0 else 0
v["pm%"] = 100 - v["ph%"] - v["pi%"] if v["pread"] > 0 else 0
v["pdhit"] = d["prefetch_data_hits"] // sint
v["pdioh"] = d["prefetch_data_iohits"] // sint
v["pdmis"] = d["prefetch_data_misses"] // sint
v["pdread"] = v["pdhit"] + v["pdioh"] + v["pdmis"]
v["pdh%"] = 100 * v["pdhit"] // v["pdread"] if v["pdread"] > 0 else 0
v["pdi%"] = 100 * v["pdioh"] // v["pdread"] if v["pdread"] > 0 else 0
v["pdm%"] = 100 - v["pdh%"] - v["pdi%"] if v["pdread"] > 0 else 0
v["pmhit"] = d["prefetch_metadata_hits"] // sint
v["pmioh"] = d["prefetch_metadata_iohits"] // sint
v["pmmis"] = d["prefetch_metadata_misses"] // sint
v["pmread"] = v["pmhit"] + v["pmioh"] + v["pmmis"]
v["pmh%"] = 100 * v["pmhit"] // v["pmread"] if v["pmread"] > 0 else 0
v["pmi%"] = 100 * v["pmioh"] // v["pmread"] if v["pmread"] > 0 else 0
v["pmm%"] = 100 - v["pmh%"] - v["pmi%"] if v["pmread"] > 0 else 0
v["mhit"] = (d["prefetch_metadata_hits"] +
d["demand_metadata_hits"]) // sint
v["mioh"] = (d["prefetch_metadata_iohits"] +
d["demand_metadata_iohits"]) // sint
v["mmis"] = (d["prefetch_metadata_misses"] +
d["demand_metadata_misses"]) // sint
v["mread"] = v["mhit"] + v["mioh"] + v["mmis"]
v["mh%"] = 100 * v["mhit"] // v["mread"] if v["mread"] > 0 else 0
v["mi%"] = 100 * v["mioh"] // v["mread"] if v["mread"] > 0 else 0
v["mm%"] = 100 - v["mh%"] - v["mi%"] if v["mread"] > 0 else 0
v["arcsz"] = cur["size"]
v["size"] = cur["size"]
v["c"] = cur["c"]
v["mfu"] = d["mfu_hits"] // sint
v["mru"] = d["mru_hits"] // sint
v["mrug"] = d["mru_ghost_hits"] // sint
v["mfug"] = d["mfu_ghost_hits"] // sint
v["unc"] = d["uncached_hits"] // sint
v["eskip"] = d["evict_skip"] // sint
v["el2skip"] = d["evict_l2_skip"] // sint
v["el2cach"] = d["evict_l2_cached"] // sint
v["el2el"] = d["evict_l2_eligible"] // sint
v["el2mfu"] = d["evict_l2_eligible_mfu"] // sint
v["el2mru"] = d["evict_l2_eligible_mru"] // sint
v["el2inel"] = d["evict_l2_ineligible"] // sint
v["mtxmis"] = d["mutex_miss"] // sint
v["ztotal"] = (d["zfetch_hits"] + d["zfetch_future"] + d["zfetch_stride"] +
d["zfetch_past"] + d["zfetch_misses"]) // sint
v["zhits"] = d["zfetch_hits"] // sint
v["zahead"] = (d["zfetch_future"] + d["zfetch_stride"]) // sint
v["zpast"] = d["zfetch_past"] // sint
v["zmisses"] = d["zfetch_misses"] // sint
v["zmax"] = d["zfetch_max_streams"] // sint
v["zfuture"] = d["zfetch_future"] // sint
v["zstride"] = d["zfetch_stride"] // sint
v["zissued"] = d["zfetch_io_issued"] // sint
v["zactive"] = d["zfetch_io_active"] // sint
if l2exist:
v["l2hits"] = d["l2_hits"] // sint
v["l2miss"] = d["l2_misses"] // sint
v["l2read"] = v["l2hits"] + v["l2miss"]
v["l2hit%"] = 100 * v["l2hits"] // v["l2read"] if v["l2read"] > 0 else 0
v["l2miss%"] = 100 - v["l2hit%"] if v["l2read"] > 0 else 0
v["l2asize"] = cur["l2_asize"]
v["l2size"] = cur["l2_size"]
v["l2bytes"] = d["l2_read_bytes"] // sint
Add L2ARC arcstats for MFU/MRU buffers and buffer content type Currently the ARC state (MFU/MRU) of cached L2ARC buffer and their content type is unknown. Knowing this information may prove beneficial in adjusting the L2ARC caching policy. This commit adds L2ARC arcstats that display the aligned size (in bytes) of L2ARC buffers according to their content type (data/metadata) and according to their ARC state (MRU/MFU or prefetch). It also expands the existing evict_l2_eligible arcstat to differentiate between MFU and MRU buffers. L2ARC caches buffers from the MRU and MFU lists of ARC. Upon caching a buffer, its ARC state (MRU/MFU) is stored in the L2 header (b_arcs_state). The l2_m{f,r}u_asize arcstats reflect the aligned size (in bytes) of L2ARC buffers according to their ARC state (based on b_arcs_state). We also account for the case where an L2ARC and ARC cached MRU or MRU_ghost buffer transitions to MFU. The l2_prefetch_asize reflects the alinged size (in bytes) of L2ARC buffers that were cached while they had the prefetch flag set in ARC. This is dynamically updated as the prefetch flag of L2ARC buffers changes. When buffers are evicted from ARC, if they are determined to be L2ARC eligible then their logical size is recorded in evict_l2_eligible_m{r,f}u arcstats according to their ARC state upon eviction. Persistent L2ARC: When committing an L2ARC buffer to a log block (L2ARC metadata) its b_arcs_state and prefetch flag is also stored. If the buffer changes its arcstate or prefetch flag this is reflected in the above arcstats. However, the L2ARC metadata cannot currently be updated to reflect this change. Example: L2ARC caches an MRU buffer. L2ARC metadata and arcstats count this as an MRU buffer. The buffer transitions to MFU. The arcstats are updated to reflect this. Upon pool re-import or on/offlining the L2ARC device the arcstats are cleared and the buffer will now be counted as an MRU buffer, as the L2ARC metadata were not updated. Bug fix: - If l2arc_noprefetch is set, arc_read_done clears the L2CACHE flag of an ARC buffer. However, prefetches may be issued in a way that arc_read_done() is bypassed. Instead, move the related code in l2arc_write_eligible() to account for those cases too. Also add a test and update manpages for l2arc_mfuonly module parameter, and update the manpages and code comments for l2arc_noprefetch. Move persist_l2arc tests to l2arc. Reviewed-by: Ryan Moeller <freqlabs@FreeBSD.org> Reviewed-by: Richard Elling <Richard.Elling@RichardElling.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: George Amanakis <gamanakis@gmail.com> Closes #10743
2020-09-14 20:10:44 +03:00
v["l2pref"] = cur["l2_prefetch_asize"]
v["l2mfu"] = cur["l2_mfu_asize"]
v["l2mru"] = cur["l2_mru_asize"]
v["l2data"] = cur["l2_bufc_data_asize"]
v["l2meta"] = cur["l2_bufc_metadata_asize"]
v["l2pref%"] = 100 * v["l2pref"] // v["l2asize"]
v["l2mfu%"] = 100 * v["l2mfu"] // v["l2asize"]
v["l2mru%"] = 100 * v["l2mru"] // v["l2asize"]
v["l2data%"] = 100 * v["l2data"] // v["l2asize"]
v["l2meta%"] = 100 * v["l2meta"] // v["l2asize"]
Add L2ARC arcstats for MFU/MRU buffers and buffer content type Currently the ARC state (MFU/MRU) of cached L2ARC buffer and their content type is unknown. Knowing this information may prove beneficial in adjusting the L2ARC caching policy. This commit adds L2ARC arcstats that display the aligned size (in bytes) of L2ARC buffers according to their content type (data/metadata) and according to their ARC state (MRU/MFU or prefetch). It also expands the existing evict_l2_eligible arcstat to differentiate between MFU and MRU buffers. L2ARC caches buffers from the MRU and MFU lists of ARC. Upon caching a buffer, its ARC state (MRU/MFU) is stored in the L2 header (b_arcs_state). The l2_m{f,r}u_asize arcstats reflect the aligned size (in bytes) of L2ARC buffers according to their ARC state (based on b_arcs_state). We also account for the case where an L2ARC and ARC cached MRU or MRU_ghost buffer transitions to MFU. The l2_prefetch_asize reflects the alinged size (in bytes) of L2ARC buffers that were cached while they had the prefetch flag set in ARC. This is dynamically updated as the prefetch flag of L2ARC buffers changes. When buffers are evicted from ARC, if they are determined to be L2ARC eligible then their logical size is recorded in evict_l2_eligible_m{r,f}u arcstats according to their ARC state upon eviction. Persistent L2ARC: When committing an L2ARC buffer to a log block (L2ARC metadata) its b_arcs_state and prefetch flag is also stored. If the buffer changes its arcstate or prefetch flag this is reflected in the above arcstats. However, the L2ARC metadata cannot currently be updated to reflect this change. Example: L2ARC caches an MRU buffer. L2ARC metadata and arcstats count this as an MRU buffer. The buffer transitions to MFU. The arcstats are updated to reflect this. Upon pool re-import or on/offlining the L2ARC device the arcstats are cleared and the buffer will now be counted as an MRU buffer, as the L2ARC metadata were not updated. Bug fix: - If l2arc_noprefetch is set, arc_read_done clears the L2CACHE flag of an ARC buffer. However, prefetches may be issued in a way that arc_read_done() is bypassed. Instead, move the related code in l2arc_write_eligible() to account for those cases too. Also add a test and update manpages for l2arc_mfuonly module parameter, and update the manpages and code comments for l2arc_noprefetch. Move persist_l2arc tests to l2arc. Reviewed-by: Ryan Moeller <freqlabs@FreeBSD.org> Reviewed-by: Richard Elling <Richard.Elling@RichardElling.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: George Amanakis <gamanakis@gmail.com> Closes #10743
2020-09-14 20:10:44 +03:00
v["grow"] = 0 if cur["arc_no_grow"] else 1
v["need"] = cur["arc_need_free"]
v["free"] = cur["memory_free_bytes"]
v["avail"] = cur["memory_available_bytes"]
Include scatter_chunk_waste in arc_size The ARC caches data in scatter ABD's, which are collections of pages, which are typically 4K. Therefore, the space used to cache each block is rounded up to a multiple of 4K. The ABD subsystem tracks this wasted memory in the `scatter_chunk_waste` kstat. However, the ARC's `size` is not aware of the memory used by this round-up, it only accounts for the size that it requested from the ABD subsystem. Therefore, the ARC is effectively using more memory than it is aware of, due to the `scatter_chunk_waste`. This impacts observability, e.g. `arcstat` will show that the ARC is using less memory than it effectively is. It also impacts how the ARC responds to memory pressure. As the amount of `scatter_chunk_waste` changes, it appears to the ARC as memory pressure, so it needs to resize `arc_c`. If the sector size (`1<<ashift`) is the same as the page size (or larger), there won't be any waste. If the (compressed) block size is relatively large compared to the page size, the amount of `scatter_chunk_waste` will be small, so the problematic effects are minimal. However, if using 512B sectors (`ashift=9`), and the (compressed) block size is small (e.g. `compression=on` with the default `volblocksize=8k` or a decreased `recordsize`), the amount of `scatter_chunk_waste` can be very large. On a production system, with `arc_size` at a constant 50% of memory, `scatter_chunk_waste` has been been observed to be 10-30% of memory. This commit adds `scatter_chunk_waste` to `arc_size`, and adds a new `waste` field to `arcstat`. As a result, the ARC's memory usage is more observable, and `arc_c` does not need to be adjusted as frequently. Reviewed-by: Pavel Zakharov <pavel.zakharov@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: George Wilson <gwilson@delphix.com> Reviewed-by: Ryan Moeller <ryan@iXsystems.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #10701
2020-08-18 06:04:04 +03:00
v["waste"] = cur["abd_chunk_waste_size"]
def main():
global sint
global count
global hdr_intr
i = 0
count_flag = 0
init()
if count > 0:
count_flag = 1
signal(SIGINT, SIG_DFL)
signal(SIGWINCH, resize_handler)
while True:
if i == 0:
print_header()
snap_stats()
calculate()
print_values()
if count_flag == 1:
if count <= 1:
break
count -= 1
i = 0 if i >= hdr_intr else i + 1
time.sleep(sint)
if out:
out.close()
if __name__ == '__main__':
main()