Permalink
Fetching contributors…
Cannot retrieve contributors at this time
184 lines (152 sloc) 9.52 KB
Demonstrations of zfsdist, the Linux eBPF/bcc version.
zfsdist traces ZFS reads, writes, opens, and fsyncs, and summarizes their
latency as a power-of-2 histogram. It has been written to work on ZFS on Linux
(http://zfsonlinux.org). For example:
# ./zfsdist
Tracing ZFS operation latency... Hit Ctrl-C to end.
^C
operation = 'read'
usecs : count distribution
0 -> 1 : 0 | |
2 -> 3 : 0 | |
4 -> 7 : 4479 |****************************************|
8 -> 15 : 1028 |********* |
16 -> 31 : 14 | |
32 -> 63 : 1 | |
64 -> 127 : 2 | |
128 -> 255 : 6 | |
256 -> 511 : 1 | |
512 -> 1023 : 1256 |*********** |
1024 -> 2047 : 9 | |
2048 -> 4095 : 1 | |
4096 -> 8191 : 2 | |
operation = 'write'
usecs : count distribution
0 -> 1 : 0 | |
2 -> 3 : 0 | |
4 -> 7 : 0 | |
8 -> 15 : 0 | |
16 -> 31 : 0 | |
32 -> 63 : 0 | |
64 -> 127 : 0 | |
128 -> 255 : 75 |****************************************|
256 -> 511 : 11 |***** |
512 -> 1023 : 0 | |
1024 -> 2047 : 0 | |
2048 -> 4095 : 0 | |
4096 -> 8191 : 0 | |
8192 -> 16383 : 0 | |
16384 -> 32767 : 0 | |
32768 -> 65535 : 0 | |
65536 -> 131071 : 13 |****** |
131072 -> 262143 : 1 | |
operation = 'open'
usecs : count distribution
0 -> 1 : 0 | |
2 -> 3 : 2 |****************************************|
This output shows a bimodal distribution for read latency, with a faster
mode of around 5 thousand reads that took between 4 and 15 microseconds, and a
slower mode of 1256 reads that took between 512 and 1023 microseconds. It's
likely that the faster mode was a hit from the in-memory file system cache,
and the slower mode is a read from a storage device (disk).
The write latency is also bimodal, with a faster mode between 128 and 511 us,
and the slower mode between 65 and 131 ms.
This "latency" is measured from when the operation was issued from the VFS
interface to the file system (via the ZFS POSIX layer), to when it completed.
This spans everything: block device I/O (disk I/O), file system CPU cycles,
file system locks, run queue latency, etc. This is a better measure of the
latency suffered by applications reading from the file system than measuring
this down at the block device interface.
Note that this only traces the common file system operations previously
listed: other file system operations (eg, inode operations including
getattr()) are not traced.
An optional interval and a count can be provided, as well as -m to show the
distributions in milliseconds. For example:
# ./zfsdist 1 5
Tracing ZFS operation latency... Hit Ctrl-C to end.
06:55:41:
operation = 'read'
usecs : count distribution
0 -> 1 : 0 | |
2 -> 3 : 0 | |
4 -> 7 : 3976 |****************************************|
8 -> 15 : 1181 |*********** |
16 -> 31 : 18 | |
32 -> 63 : 4 | |
64 -> 127 : 17 | |
128 -> 255 : 16 | |
256 -> 511 : 0 | |
512 -> 1023 : 1275 |************ |
1024 -> 2047 : 36 | |
2048 -> 4095 : 3 | |
4096 -> 8191 : 0 | |
8192 -> 16383 : 1 | |
16384 -> 32767 : 1 | |
06:55:42:
operation = 'read'
usecs : count distribution
0 -> 1 : 0 | |
2 -> 3 : 0 | |
4 -> 7 : 12751 |****************************************|
8 -> 15 : 1190 |*** |
16 -> 31 : 38 | |
32 -> 63 : 7 | |
64 -> 127 : 85 | |
128 -> 255 : 47 | |
256 -> 511 : 0 | |
512 -> 1023 : 1010 |*** |
1024 -> 2047 : 49 | |
2048 -> 4095 : 12 | |
06:55:43:
operation = 'read'
usecs : count distribution
0 -> 1 : 0 | |
2 -> 3 : 0 | |
4 -> 7 : 80925 |****************************************|
8 -> 15 : 1645 | |
16 -> 31 : 251 | |
32 -> 63 : 24 | |
64 -> 127 : 16 | |
128 -> 255 : 12 | |
256 -> 511 : 0 | |
512 -> 1023 : 80 | |
1024 -> 2047 : 1 | |
06:55:44:
operation = 'read'
usecs : count distribution
0 -> 1 : 0 | |
2 -> 3 : 0 | |
4 -> 7 : 81207 |****************************************|
8 -> 15 : 2075 |* |
16 -> 31 : 2005 | |
32 -> 63 : 177 | |
64 -> 127 : 3 | |
06:55:45:
operation = 'read'
usecs : count distribution
0 -> 1 : 0 | |
2 -> 3 : 0 | |
4 -> 7 : 74364 |****************************************|
8 -> 15 : 865 | |
16 -> 31 : 4960 |** |
32 -> 63 : 625 | |
64 -> 127 : 2 | |
This workload was randomly reading from a file that became cached. The slower
mode can be seen to disappear by the final summaries.
USAGE message:
# ./zfsdist -h
usage: zfsdist [-h] [-T] [-m] [-p PID] [interval] [count]
Summarize ZFS operation latency
positional arguments:
interval output interval, in seconds
count number of outputs
optional arguments:
-h, --help show this help message and exit
-T, --notimestamp don't include timestamp on interval output
-m, --milliseconds output in milliseconds
-p PID, --pid PID trace this PID only
examples:
./zfsdist # show operation latency as a histogram
./zfsdist -p 181 # trace PID 181 only
./zfsdist 1 10 # print 1 second summaries, 10 times
./zfsdist -m 5 # 5s summaries, milliseconds