Ceph分布式集群文件系统测试
1、使用Ceph rados bech工具进行测试rados bech工具是Ceph内置的基准性能测试工具，用于测试Ceph集群存储池层面的性能。rados bech支持写、连续读、和随机读等 基准测试。
rados bech工具命令语法：
#rados bench -p <pool name> <seconds> <write|seq|rand> -b <blcoksize> -t <> --no-cleanup
-p：pool存储池名称
<second>：测试时间，单位秒
<write|seq|rand>：写、顺序读、随机读
-b：块大小，默认是4秒
-t：并发线程数，默认是16
--no-cleanup：临时写入存储池中的数据不被清除，默认会清除
在不清除数据的情况下，进行1000秒的写测试：
# rados bench -p pool1 1000 write --no-cleanup
Total time run: 1003.149409
Total writes made: 6627
Write size: 4194304
Object size: 4194304
Bandwidth (MB/sec): 26.4248
Stddev Bandwidth: 13.4511
Max bandwidth (MB/sec): 116
Min bandwidth (MB/sec): 0
Average IOPS: 6
Stddev IOPS: 3
Max IOPS: 29
Min IOPS: 0
Average Latency(s): 2.42078
Stddev Latency(s): 0.951884
Max latency(s): 6.59119
Min latency(s): 0.119255
使用ceph df查看到集群空间使用情况：
# ceph df
GLOBAL:
SIZE AVAIL RAW USED %RAW USED
40966G 40555G 411G 1.00
POOLS:
NAME ID USED %USED MAX AVAIL OBJECTS
rbd 0 6168M 0.03 19357G 1543
pool1 1 89872M 0.43 19357G 22469
pool2 2 111G 0.54 19357G 28510
对存储池进行300秒顺序读测试
# rados bench -p pool1 300 seq
Total time run: 125.494596
Total reads made: 6627
Read size: 4194304
Object size: 4194304
Bandwidth (MB/sec): 211.228
Average IOPS 52
Stddev IOPS: 8
Max IOPS: 73
Min IOPS: 29
Average Latency(s): 0.301849
Max latency(s): 2.17986
Min latency(s): 0.0179132
对pool1存储池进行随机读测试，测试时间为10分钟
1、
# rados bench -p rbd 600 rand
2016-09-22 22:19:41.815759 min lat: 0.0178813 max lat: 2.69039 avg lat: 0.33182
sec Cur ops startedfinishedavg MB/scur MB/s last lat(s)avg lat(s)
600 16 28860 28844 192.274 192 0.511941 0.33182
Total time run: 600.544174
Total reads made: 28860
Read size: 4194304
Object size: 4194304
Bandwidth (MB/sec): 192.226
Average IOPS: 48
Stddev IOPS: 7
Max IOPS: 77
Min IOPS: 19
Average Latency(s): 0.332086
Max latency(s): 2.69039
Min latency(s): 0.0178813
使用nmon查看disk读写情况
使用watch ceph -s查看集群读写情况
2、rado load-gen工具
rados load-gen用于生成Ceph集群负载，也可用于高负载下模拟测试
参数说明如下：
rados -p rbd load-gen \
--num-objects 50 \
#对象总数
--min-objects-size 4M \
#对象最小字节数
--max-objects-size 4M \
#对象最大字节数
--max-ops 16 \
#操作最大数
--min-ops \
#操作最小数
--min-op-len 4M \
#操作最小长度
--max-op-len 4M \
#操作最大长度
--percent 5 \
#读操作百分比，读写混合中读的比例
--target-throughput 2000 \
目标吞吐量，单位MB
--run-length 60
运行时间，单位秒
# cat rados-load-gen.sh
#!/bin/bash
rados -p rbd load-gen \
--num-objects 50 \
--min-objects-size 4M \
--max-objects-size 4M \
--max-ops 16 \
--min-op-len 4M \
--max-op-len 4M \
--percent 5 \
--target-throughput 2000 \
--run-length 60
命令会通过将50个对像写到rbd存储池来生成负载，每个对象大小为4MB，读写比例为5%,总的运行时间为60秒
测试：
# sh rados-load-gen.sh
57: throughput=55.6MB/sec pending data=8388608
READ : oid=obj-pR0IiIQUAhoG2vr off=539347588 len=4194304
op 829 completed, throughput=55.6MB/sec
READ : oid=obj-QJuP_GXN91TZANt off=894509182 len=4194304
op 831 completed, throughput=55.5MB/sec
READ : oid=obj-FSNdJlg0iLzWeR7 off=1550916982 len=4194304
op 832 completed, throughput=55.6MB/sec
READ : oid=obj-pqhOi8w3ChbnMSF off=302850550 len=4194304
op 833 completed, throughput=55.5MB/sec
waiting for all operations to complete
op 834 completed, throughput=55.5MB/sec
cleaning up objects
op 773 completed, throughput=29.9MB/sec
3、Ceph块设备基准测试
# cat rdb-bench-write.sh
#!/bin/bash
#1、创建CEPH块设备，名为block-device,size为1GB，并映射它：
rbd create block-device --size 10240
rbd info --image block-device
rbd map block-device
rbd showmapped
#2、在块设备上创建文件系统并挂载它：
mkfs.xfs /dev/rbd0
mkdir -p /mnt/ceph-block-device
mount /dev/rbd0 /mnt/ceph-block-device
df -h /mnt/ceph-block/device
#3、对blcok-device进行5GB数据写基准测试：
rbd bench-write block-device --io-total 5368709200
rdb bench-write语法格式
#rdb bech-write <RBD Image name>
--io-size:写入字节，默认为4MB
--io-threads:线程数，默认为16
--io-total：写入总字节，默认为1024MB
--io-pattern <seq/rand>：写模式，默认为顺序写
创建Ceph块设备
# rbd create block-device1 --size 1G --image-format 1
rbd: image format 1 is deprecated
# rbd info --image block-device1
rbd image 'block-device1':
size 1024 MB in 256 objects
order 22 (4096 kB objects)
block_name_prefix: rb.0.19d5.238e1f29
format: 1
# rbd map block-device1
/dev/rbd0
# rbd showmapped
id pool image snap device
0rbdblock-device1 - /dev/rbd0
在块设备上创建文件系统并挂载它：
# mkfs.xfs /dev/rbd0
meta-data=/dev/rbd0 isize=256 agcount=9, agsize=31744 blks
= sectsz=512 attr=2, projid32bit=1
= crc=0 finobt=0
data = bsize=4096 blocks=262144, imaxpct=25
= sunit=1024 swidth=1024 blks
naming =version 2 bsize=4096 ascii-ci=0 ftype=0
log =internal log bsize=4096 blocks=2560, version=2
= sectsz=512 sunit=8 blks, lazy-count=1
realtime =none extsz=4096 blocks=0, rtextents=0
# mkdir -p /mnt/ceph-block-device1
# mount /dev/rbd0 /mnt/ceph-block-device
ceph-block-device/ceph-block-device1/
# mount /dev/rbd0 /mnt/ceph-block-device1
# df -h /mnt/ceph-block-device1
Filesystem SizeUsed Avail Use% Mounted on
/dev/rbd0 1014M 33M982M 4% /mnt/ceph-block-device1
对blcok-device进行5GB数据写基准测试：
rand:随机写，seq顺序写
# rbd bench-write block-device1 --io-total 5368709200 --io-pattern rand
elapsed: 488ops:1310721ops/sec:2681.48bytes/sec: 10983362.38
# rbd bench-write block-device1 --io-total 5368709200 --io-pattern seq
elapsed: 170ops:1310721ops/sec:7701.13bytes/sec: 31543830.94
4、使用FIO做Ceph rbd基准测试
测试方法：
先创建一个块设备，映射到ceph客户端节点。前面测试已经创建block-device1，使用它进行测试
安装下fio，使用yum install fio，fio支持RBD IO引擎，我们需要提供RBD映像名称、存储池用用于连接ceph集群的用户名就可以，创建下面fio配置文件：
rw=write，进行顺序写测试
# vim write.fio
description="write test with block size of 4M"
ioengine=rbd
clientname=admin
pool=rbd
rbdname=block-device1
iodepth=32
runtime=120
rw=write
bs=4M
Last login: Fri Sep 23 02:54:12 2016 from 192.168.13.101
# fio write.fio
write-4M: (g=0): rw=write, bs=4M-4M/4M-4M/4M-4M, ioengine=rbd, iodepth=32
fio-2.3
time 15258 cycles_start=124103204639059
Starting 1 process
rbd engine: RBD version: 0.1.10
Jobs: 1 (f=1): Jobs: 1 (f=1): Jobs: 1 (f=1): Jobs: 1 (f=1):
write-4M: (groupid=0, jobs=1): err= 0: pid=15285: Fri Sep 23 02:58:55 2016
Description: ["write test with block size of 4M"]
write: io=1024.0MB, bw=64365KB/s, iops=15, runt= 16291msec
slat (usec): min=70, max=527, avg=239.13, stdev=86.36
clat (msec): min=953, max=4987, avg=2000.13, stdev=806.32
lat (msec): min=953, max=4987, avg=2000.37, stdev=806.32
clat percentiles (msec):
|1.00th=,5.00th=, 10.00th=[ 1139], 20.00th=[ 1188],
| 30.00th=[ 1483], 40.00th=[ 1827], 50.00th=[ 1827], 60.00th=[ 2008],
| 70.00th=[ 2114], 80.00th=[ 2868], 90.00th=[ 3195], 95.00th=[ 3523],
| 99.00th=[ 5014], 99.50th=[ 5014], 99.90th=[ 5014], 99.95th=[ 5014],
| 99.99th=[ 5014]
bw (KB/s): min= 2773, max=103290, per=98.03%, avg=63095.40, stdev=30372.18
lat (msec) : 1000=8.98%, 2000=44.92%, >=2000=46.09%
cpu : usr=0.38%, sys=0.00%, ctx=14, majf=0, minf=0
IO depths : 1=0.8%, 2=1.6%, 4=5.9%, 8=26.6%, 16=61.3%, 32=3.9%, >=64=0.0%
submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
complete: 0=0.0%, 4=96.3%, 8=0.0%, 16=0.0%, 32=3.7%, 64=0.0%, >=64=0.0%
issued : total=r=0/w=256/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0
latency : target=0, window=0, percentile=100.00%, depth=32
Run status group 0 (all jobs):
WRITE: io=1024.0MB, aggrb=64365KB/s, minb=64365KB/s, maxb=64365KB/s, mint=16291msec, maxt=16291msec
Disk stats (read/write):
sda: ios=0/6, merge=0/0, ticks=0/100, in_queue=100, util=0.31%
rw=read，进行读测试
# vim write.fio
description="write test with block size of 4M"
ioengine=rbd
clientname=admin
pool=rbd
rbdname=block-device1
iodepth=32
runtime=120
rw=read
bs=4M
# fio write.fio
write-4M: (g=0): rw=read, bs=4M-4M/4M-4M/4M-4M, ioengine=rbd, iodepth=32
fio-2.3
time 15383 cycles_start=124168557431260
Starting 1 process
rbd engine: RBD version: 0.1.10
Jobs: 1 (f=1):
write-4M: (groupid=0, jobs=1): err= 0: pid=15410: Fri Sep 23 03:00:52 2016
Description: ["write test with block size of 4M"]
read : io=1024.0MB, bw=198745KB/s, iops=48, runt=5276msec
slat (usec): min=3, max=17, avg= 4.19, stdev= 2.61
clat (msec): min=162, max=1339, avg=640.89, stdev=201.60
lat (msec): min=162, max=1339, avg=640.90, stdev=201.60
clat percentiles (msec):
|1.00th=,5.00th=, 10.00th=, 20.00th=,
| 30.00th=, 40.00th=, 50.00th=, 60.00th=,
| 70.00th=, 80.00th=, 90.00th=, 95.00th=[ 1090],
| 99.00th=[ 1270], 99.50th=[ 1270], 99.90th=[ 1336], 99.95th=[ 1336],
| 99.99th=[ 1336]
bw (KB/s): min= 5927, max=244017, per=86.44%, avg=171799.00, stdev=83676.59
lat (msec) : 250=1.17%, 500=33.20%, 750=51.95%, 1000=7.42%, 2000=6.25%
cpu : usr=0.06%, sys=0.00%, ctx=18, majf=0, minf=0
IO depths : 1=0.8%, 2=3.1%, 4=8.2%, 8=28.1%, 16=56.2%, 32=3.5%, >=64=0.0%
submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
complete: 0=0.0%, 4=96.6%, 8=0.0%, 16=0.0%, 32=3.4%, 64=0.0%, >=64=0.0%
issued : total=r=256/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0
latency : target=0, window=0, percentile=100.00%, depth=32
Run status group 0 (all jobs):
READ: io=1024.0MB, aggrb=198744KB/s, minb=198744KB/s, maxb=198744KB/s, mint=5276msec, maxt=5276msec
Disk stats (read/write):
sda: ios=0/0, merge=0/0, ticks=0/0, in_queue=0, util=0.00%
rw=randread，进行随机读测试
# cat write.fio
description="write test with block size of 4M"
ioengine=rbd
clientname=admin
pool=rbd
rbdname=block-device1
iodepth=32
runtime=120
rw=randread
bs=4M
# fio write.fio
write-4M: (g=0): rw=randread, bs=4M-4M/4M-4M/4M-4M, ioengine=rbd, iodepth=32
fio-2.3
time 16722 cycles_start=124259845112132
Starting 1 process
rbd engine: RBD version: 0.1.10
Jobs: 1 (f=1):
write-4M: (groupid=0, jobs=1): err= 0: pid=16749: Fri Sep 23 03:03:50 2016
Description: ["write test with block size of 4M"]
read : io=1024.0MB, bw=194181KB/s, iops=47, runt=5400msec
slat (usec): min=3, max=20, avg= 4.32, stdev= 2.58
clat (msec): min=359, max=1491, avg=650.89, stdev=174.32
lat (msec): min=359, max=1491, avg=650.90, stdev=174.32
clat percentiles (msec):
|1.00th=,5.00th=, 10.00th=, 20.00th=,
| 30.00th=, 40.00th=, 50.00th=, 60.00th=,
| 70.00th=, 80.00th=, 90.00th=, 95.00th=,
| 99.00th=[ 1450], 99.50th=[ 1500], 99.90th=[ 1500], 99.95th=[ 1500],
| 99.99th=[ 1500]
bw (KB/s): min=102767, max=318950, per=98.17%, avg=190635.00, stdev=65800.97
lat (msec) : 500=23.83%, 750=48.83%, 1000=23.83%, 2000=3.52%
cpu : usr=0.06%, sys=0.00%, ctx=15, majf=0, minf=0
IO depths : 1=1.6%, 2=5.1%, 4=12.1%, 8=28.1%, 16=50.0%, 32=3.1%, >=64=0.0%
submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
complete: 0=0.0%, 4=96.6%, 8=0.0%, 16=0.4%, 32=3.0%, 64=0.0%, >=64=0.0%
issued : total=r=256/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0
latency : target=0, window=0, percentile=100.00%, depth=32
Run status group 0 (all jobs):
READ: io=1024.0MB, aggrb=194180KB/s, minb=194180KB/s, maxb=194180KB/s, mint=5400msec, maxt=5400msec
Disk stats (read/write):
sda: ios=0/2, merge=0/0, ticks=0/20, in_queue=20, util=0.37%
#
rw= randwrite，进行随机写测试
# cat write.fio
description="write test with block size of 4M"
ioengine=rbd
clientname=admin
pool=rbd
rbdname=block-device1
iodepth=32
runtime=120
rw=randwrite
bs=4M
# fio write.fio
write-4M: (g=0): rw=randwrite, bs=4M-4M/4M-4M/4M-4M, ioengine=rbd, iodepth=32
fio-2.3
time 16834 cycles_start=124288020352425
Starting 1 process
rbd engine: RBD version: 0.1.10
Jobs: 1 (f=1): Jobs: 1 (f=1): Jobs: 1 (f=1): Jobs: 1 (f=1): Jobs: 1 (f=1):
write-4M: (groupid=0, jobs=1): err= 0: pid=16861: Fri Sep 23 03:04:57 2016
Description: ["write test with block size of 4M"]
write: io=1024.0MB, bw=63465KB/s, iops=15, runt= 16522msec
slat (usec): min=89, max=508, avg=236.04, stdev=85.71
clat (msec): min=1059, max=4915, avg=2038.61, stdev=762.83
lat (msec): min=1060, max=4916, avg=2038.84, stdev=762.83
clat percentiles (msec):
|1.00th=[ 1057],5.00th=[ 1057], 10.00th=[ 1074], 20.00th=[ 1254],
| 30.00th=[ 1270], 40.00th=[ 1975], 50.00th=[ 2008], 60.00th=[ 2343],
| 70.00th=[ 2442], 80.00th=[ 2474], 90.00th=[ 2933], 95.00th=[ 2933],
| 99.00th=[ 4883], 99.50th=[ 4948], 99.90th=[ 4948], 99.95th=[ 4948],
| 99.99th=[ 4948]
bw (KB/s): min= 3202, max=110909, per=95.10%, avg=60356.44, stdev=30229.78
lat (msec) : 2000=41.41%, >=2000=58.59%
cpu : usr=0.38%, sys=0.00%, ctx=11, majf=0, minf=0
IO depths : 1=0.8%, 2=3.1%, 4=9.8%, 8=26.2%, 16=55.9%, 32=4.3%, >=64=0.0%
submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
complete: 0=0.0%, 4=96.6%, 8=0.0%, 16=0.0%, 32=3.4%, 64=0.0%, >=64=0.0%
issued : total=r=0/w=256/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0
latency : target=0, window=0, percentile=100.00%, depth=32
Run status group 0 (all jobs):
WRITE: io=1024.0MB, aggrb=63465KB/s, minb=63465KB/s, maxb=63465KB/s, mint=16522msec, maxt=16522msec
Disk stats (read/write):
sda: ios=0/3, merge=0/0, ticks=0/20, in_queue=20, util=0.12%
5、单个磁盘性能测试
# ceho 3 > /proc/sys/vm/drop_caches
# dd if=/dev/zero of=/var/lib/ceph/osd/ceph-12/deleteme bs=10G count=1 oflag=direct
0+1 records in
0+1 records out
2147418112 bytes (2.1 GB) copied, 13.0371 s, 165 MB/s
#
单个磁盘读性能
# dd if=/var/lib/ceph/osd/ceph-12/deleteme of=/dev/null bs=10G count=1 iflag=direct
0+1 records in
0+1 records out
2147418112 bytes (2.1 GB) copied, 11.8353 s, 181 MB/s
6、使用Iperf进行ceph节点之间的网络性能，
A,在第一个ceph节点执行iperf服务器端选项，带-s 参数并使用-p选项指定端口进行监听。
# iperf -s -p 6900
------------------------------------------------------------
Server listening on TCP port 6900
TCP window size: 85.3 KByte (default)
------------------------------------------------------------
local 10.1.1.30 port 6900 connected with 10.1.1.20 port 45437
[ ID] Interval Transfer Bandwidth
0.0-10.0 sec10.7 GBytes9.17 Gbits/sec
B,在另一个ceph节点使用-c参数以客户端选项运行，并接server端节点名
# iperf -c node3 -p 6900
------------------------------------------------------------
Client connecting to node3, TCP port 6900
TCP window size: 85.0 KByte (default)
------------------------------------------------------------
local 10.1.1.20 port 45437 connected with 10.1.1.30 port 6900
[ ID] Interval Transfer Bandwidth
0.0-10.0 sec10.7 GBytes9.17 Gbits/sec
页:
[1]