Skip to content

B
bcc
Commit 2b348704 authored by Joel Fernandes's avatar Joel Fernandes
Browse files
Options

tools/cachestat: Rewrite it so it makes more sense 

cachstat separation between read and writes doesn't make much sense. During tests
its seen that a workload that is supposed to be completely in the missed category
appears to be 50% hit.

This patch rewrites the BCC tool to be more inline with Brendan Gregg's original
cachestat perf tool which makes more sense:
https://github.com/brendangregg/perf-tools/blob/master/fs/cachestat
http://www.brendangregg.com/blog/2014-12-31/linux-page-cache-hit-ratio.html

Fixes Issue: #1586
Signed-off-by: default avatarJoel Fernandes <joelaf@google.com>
parent cab4d7c6
Show whitespace changes
Inline Side-by-side
Showing with 66 additions and 91 deletions
tools/cachestat.py
View file @ 2b348704
...@@ -39,16 +39,13 @@ def get_meminfo(): ...@@ -39,16 +39,13 @@ def get_meminfo():
return result return result
# set global variables # set global variables
rtaccess = 0
wtaccess = 0
mpa = 0 mpa = 0
mbd = 0 mbd = 0
apcl = 0 apcl = 0
apd = 0 apd = 0
access = 0 total = 0
misses = 0 misses = 0
rhits = 0 hits = 0
whits = 0
debug = 0 debug = 0
# args # args
...@@ -116,9 +113,8 @@ b.attach_kprobe(event="mark_buffer_dirty", fn_name="do_count") ...@@ -116,9 +113,8 @@ b.attach_kprobe(event="mark_buffer_dirty", fn_name="do_count")
# header # header
if tstamp: if tstamp:
print("%-8s " % "TIME", end="") print("%-8s " % "TIME", end="")
print("%8s %8s %8s %10s %10s %12s %10s" % print("%8s %8s %8s %8s %12s %10s" %
("HITS", "MISSES", "DIRTIES", ("TOTAL", "MISSES", "HITS", "DIRTIES", "BUFFERS_MB", "CACHED_MB"))
"READ_HIT%", "WRITE_HIT%", "BUFFERS_MB", "CACHED_MB"))
loop = 0 loop = 0
exiting = 0 exiting = 0
...@@ -150,28 +146,29 @@ while 1: ...@@ -150,28 +146,29 @@ while 1:
if re.match(b'account_page_dirtied', b.ksym(k.ip)) is not None: if re.match(b'account_page_dirtied', b.ksym(k.ip)) is not None:
apd = max(0, v.value) apd = max(0, v.value)
# access = total cache access incl. reads(mpa) and writes(mbd) # total = total cache accesses without counting dirties
# misses = total of add to lru which we do when we write(mbd) # misses = total of add to lru because of read misses
# and also the mark the page dirty(same as mbd) total = (mpa - mbd)
access = (mpa + mbd) misses = (apcl - apd)
misses = (apcl + apd)
# rtaccess is the read hit % during the sample period. if total < 0:
# wtaccess is the write hit % during the smaple period. total = 0
if mpa > 0:
rtaccess = float(mpa) / (access + misses)
if apcl > 0:
wtaccess = float(apcl) / (access + misses)
if wtaccess != 0: if misses < 0:
whits = 100 * wtaccess misses = 0
if rtaccess != 0:
rhits = 100 * rtaccess hits = total - misses
# If hits are < 0, then its possible misses are overestimated
# due to possibly page cache read ahead adding more pages than
# needed. In this case just assume misses as total and reset hits.
if hits < 0:
misses = total
hits = 0
if debug: if debug:
print("%d %d %d %d %d %d %f %f %d %d\n" % print("%d %d %d %d %d %d %d\n" %
(mpa, mbd, apcl, apd, access, misses, (mpa, mbd, apcl, apd, total, misses, hits))
rtaccess, wtaccess, rhits, whits))
counts.clear() counts.clear()
...@@ -182,21 +179,18 @@ while 1: ...@@ -182,21 +179,18 @@ while 1:
if tstamp == 1: if tstamp == 1:
print("%-8s " % strftime("%H:%M:%S"), end="") print("%-8s " % strftime("%H:%M:%S"), end="")
print("%8d %8d %8d %9.1f%% %9.1f%% %12.0f %10.0f" % print("%8d %8d %8d %8d %12.0f %10.0f" %
(access, misses, mbd, rhits, whits, buff, cached)) (total, misses, hits, mbd, buff, cached))
mpa = 0 mpa = 0
mbd = 0 mbd = 0
apcl = 0 apcl = 0
apd = 0 apd = 0
access = 0 total = 0
misses = 0 misses = 0
rhits = 0 hits = 0
cached = 0 cached = 0
buff = 0 buff = 0
whits = 0
rtaccess = 0
wtaccess = 0
if exiting: if exiting:
print("Detaching...") print("Detaching...")
......
tools/cachestat_example.txt
View file @ 2b348704
...@@ -15,61 +15,42 @@ examples: ...@@ -15,61 +15,42 @@ examples:
./cachestat -T 1 5 # include timestamps with interval of one second for five iterations ./cachestat -T 1 5 # include timestamps with interval of one second for five iterations
# ./cachestat 1 Following commands show a 2GB file being read into the page cache.
HITS MISSES DIRTIES READ_HIT% WRITE_HIT% BUFFERS_MB CACHED_MB
0 58 0 0.0% 100.0% 0 11334 Command used to generate activity:
146113 0 0 100.0% 0.0% 0 11334 # dd if=/root/tmpfile of=/dev/null bs=8192
244143 0 0 100.0% 0.0% 0 11334
216833 0 0 100.0% 0.0% 0 11334 Output from cachestat running simultatenously:
248209 0 0 100.0% 0.0% 0 11334 # ./tools/cachestat.py 1
205825 0 0 100.0% 0.0% 0 11334 TOTAL MISSES HITS DIRTIES BUFFERS_MB CACHED_MB
286654 0 0 100.0% 0.0% 0 11334 1 0 1 0 8 283
275850 0 0 100.0% 0.0% 0 11334 0 0 0 0 8 283
272883 0 0 100.0% 0.0% 0 11334 0 0 0 2 8 283
261633 0 0 100.0% 0.0% 0 11334 0 0 0 0 8 283
252826 0 0 100.0% 0.0% 0 11334 10009 9173 836 2 9 369
235253 70 3 100.0% 0.0% 0 11335 152032 152032 0 0 9 1028
204946 0 0 100.0% 0.0% 0 11335 157408 157405 3 0 9 1707
0 0 0 0.0% 0.0% 0 11335 150432 150432 0 0 9 2331
0 0 0 0.0% 0.0% 0 11335 0 0 0 0 9 2331
0 0 0 0.0% 0.0% 0 11335 1 1 0 1 9 2331
0 0 0 0 9 2331
Above shows the reading of a 12GB file already cached in the OS page cache and again below with timestamps. 0 0 0 0 9 2331
0 0 0 0 9 2331
# ./cachestat -T 1
TIME HITS MISSES DIRTIES READ_HIT% WRITE_HIT% BUFFERS_MB CACHED_MB The misses counter reflects a 2GB file being read and almost everything being
16:07:10 0 0 0 0.0% 0.0% 0 11336 a page cache miss.
16:07:11 0 0 0 0.0% 0.0% 0 11336
16:07:12 117849 0 0 100.0% 0.0% 0 11336 Below shows an example of a new 100MB file added to page cache, by using
16:07:13 212558 0 0 100.0% 0.0% 0 11336 the command: dd if=/dev/zero of=/root/tmpfile2 bs=4k count=$((256*100))
16:07:14 302559 1 0 100.0% 0.0% 0 11336
16:07:15 309230 0 0 100.0% 0.0% 0 11336 # ./tools/cachestat.py 1
16:07:16 305701 0 0 100.0% 0.0% 0 11336 TOTAL MISSES HITS DIRTIES BUFFERS_MB CACHED_MB
16:07:17 312754 0 0 100.0% 0.0% 0 11336 0 0 0 0 15 2440
16:07:18 308406 0 0 100.0% 0.0% 0 11336 0 0 0 0 15 2440
16:07:19 298185 0 0 100.0% 0.0% 0 11336 0 0 0 0 15 2440
16:07:20 236128 0 0 100.0% 0.0% 0 11336 1758 0 1758 25603 15 2540
16:07:21 257616 0 0 100.0% 0.0% 0 11336 0 0 0 0 15 2540
16:07:22 179792 0 0 100.0% 0.0% 0 11336 0 0 0 0 15 2541
Command used to generate the activity ~25600 pages are being dirtied (writes) which corresponds to the 100MB file
# dd if=/root/mnt2/testfile of=/dev/null bs=8192 added to the page cache.
1442795+0 records in
1442795+0 records out
11819376640 bytes (12 GB) copied, 3.9301 s, 3.0 GB/s
Below shows the dirty ratio increasing as we add a file to the page cache
# ./cachestat.py
HITS MISSES DIRTIES READ_HIT% WRITE_HIT% BUFFERS_MB CACHED_MB
1074 44 13 94.9% 2.9% 1 223
2195 170 8 92.5% 6.8% 1 143
182 53 56 53.6% 1.3% 1 143
62480 40960 20480 40.6% 19.8% 1 223
7 2 5 22.2% 22.2% 1 223
348 0 0 100.0% 0.0% 1 223
0 0 0 0.0% 0.0% 1 223
0 0 0 0.0% 0.0% 1 223
0 0 0 0.0% 0.0% 1 223
0 0 0 0.0% 0.0% 1 223
The file copied into page cache was named 80m with a size of 83886080 (83886080/4096) = 20480, this matches what we see under dirties
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment
GitLab Nexedi Edition | About GitLab | About Nexedi | 沪ICP备2021021310号-2 | 沪ICP备2021021310号-7