├── .gitignore ├── gvp.sh ├── gvp-client.sh ├── README.md ├── gvp-README.html ├── LICENSE ├── extract-gl-client-prof.py └── extract-glvolprof.py /.gitignore: -------------------------------------------------------------------------------- 1 | # Byte-compiled / optimized / DLL files 2 | __pycache__/ 3 | *.py[cod] 4 | 5 | # C extensions 6 | *.so 7 | 8 | # Distribution / packaging 9 | .Python 10 | env/ 11 | build/ 12 | develop-eggs/ 13 | dist/ 14 | downloads/ 15 | eggs/ 16 | .eggs/ 17 | lib/ 18 | lib64/ 19 | parts/ 20 | sdist/ 21 | var/ 22 | *.egg-info/ 23 | .installed.cfg 24 | *.egg 25 | 26 | # PyInstaller 27 | # Usually these files are written by a python script from a template 28 | # before PyInstaller builds the exe, so as to inject date/other infos into it. 29 | *.manifest 30 | *.spec 31 | 32 | # Installer logs 33 | pip-log.txt 34 | pip-delete-this-directory.txt 35 | 36 | # Unit test / coverage reports 37 | htmlcov/ 38 | .tox/ 39 | .coverage 40 | .coverage.* 41 | .cache 42 | nosetests.xml 43 | coverage.xml 44 | *,cover 45 | 46 | # Translations 47 | *.mo 48 | *.pot 49 | 50 | # Django stuff: 51 | *.log 52 | 53 | # Sphinx documentation 54 | docs/_build/ 55 | 56 | # PyBuilder 57 | target/ 58 | -------------------------------------------------------------------------------- /gvp.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # gvp.sh - collect performance data from Gluster about a particular Gluster volume 3 | # usage: 4 | # bash gvp.sh your-gluster-volume sample-count sample-interval output-file 5 | # output-file is optional, defaults to gvp.log 6 | # 7 | # this version of the script puts the data in pbench format 8 | # for HTML graph generation 9 | # 10 | volume_name=$1 11 | sample_count=$2 12 | sample_interval=$3 13 | outfile=$4 14 | if [ "$sample_interval" = "" ] ; then 15 | echo "usage: gvp.sh your-gluster-volume sample-count sample-interval-sec [ output-file ] " 16 | exit 1 17 | fi 18 | if [ -z "$outfile" ] ; then outfile=gvp.log ; fi 19 | 20 | # start up profiling 21 | 22 | gluster volume profile $volume_name start 23 | gluster volume profile $volume_name info > /tmp/past 24 | 25 | # record a timestamp so we know when the data was collected 26 | # this lets us generate timestamps to put .csv output in pbench format 27 | 28 | date +%Y-%m-%d-%H-%M > $outfile 29 | echo "$sample_interval $sample_count" >> $outfile 30 | 31 | # generate samples 32 | for min in `seq 1 $sample_count` ; do 33 | sleep $sample_interval 34 | gluster volume profile $volume_name info 35 | done >> $outfile 36 | gluster volume profile $volume_name stop 37 | echo "output written to $outfile" 38 | -------------------------------------------------------------------------------- /gvp-client.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # gvp-client.sh - collect perf data from Gluster for client's usage 3 | # of Gluster volume from 1 mountpoint 4 | # 5 | # ASSUMPTION: "gluster volume profile your-volume start" has already happened 6 | # 7 | # usage: 8 | # chmod u+x gvp-client.sh 9 | # ./gvp-client.sh your-gluster-volume your-client-mountpoint samples interval [ logfile ] 10 | # 11 | volume_name=$1 12 | mountpoint=$2 13 | sample_count=$3 14 | sample_interval=$4 15 | your_logfile=$5 16 | if [ "$sample_interval" = "" ] ; then 17 | echo "usage: gvp-client.sh your-gluster-volume your-client-mountpoint sample-count sample-interval-sec [ your-logfile ]" 18 | exit 1 19 | fi 20 | 21 | sample_cmd="setfattr -n trusted.io-stats-dump -v " 22 | 23 | timestamp=`date +%Y-%m-%d-%H-%M` 24 | logfile=/var/tmp/gvp-client-${timestamp}.log 25 | if [ -n "$your_logfile" ] ; then 26 | logfile="$your_logfile" 27 | fi 28 | 29 | # make sure not polluted with previous data 30 | rm -f $logfile 31 | 32 | # so next sample interval will be $sample_interval 33 | $sample_cmd /var/tmp/gvp.log $mountpoint 34 | 35 | for min in `seq 1 $sample_count` ; do 36 | sleep $sample_interval 37 | rm -f /var/tmp/gvp.log 38 | $sample_cmd /var/tmp/gvp.log $mountpoint 39 | ( date ; cat /var/tmp/gvp.log ) >> $logfile 40 | done 41 | echo "output written to $logfile" 42 | 43 | -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | ## Introduction 2 | 3 | The extract-glvolprof.py program is meant to assist with visualizing the performance of 4 | a gluster volume, using the gluster volume profile command. One of key concepts in Gluster is the FOP (File Operation). This is the unit of work passed from the application down through the Gluster translator stack until it reaches the storage device. FOP Types for file creation, reading, writing, and many others are observable with these tools. 5 | 6 | 7 | Statistic types produced per FOP type by these scripts include: 8 | 9 | - call rates - for example, how many requests of different types are made per sec 10 | - % latency - what fraction of FOP response time is consumed by different FOP types 11 | - avg. latency - average FOP response time 12 | - minimum latency 13 | - maximum latency 14 | 15 | Where all latencies are in units of microseconds. 16 | 17 | The profiling tools consist of a collection and extraction script. Typically you run the collection script to collect the profile data on a Gluster client or server, and then copy the file to your local system to run the extraction tool, which is just a python text processing script and should run anywhere. 18 | 19 | To install, after cloning this repo, install the pbench-web-server RPM, which contains some javascript libraries that are used by the HTML file above and provide common code to read CSV files and produce graphs using the nvd3 library. This code comes from the pbench project at: 20 | 21 | https://github.com/distributed-system-analysis/pbench 22 | 23 | These tools produce a subdirectory containing java-script graphs that can be viewed with a web browser, as well as .csv-format files that can be loaded into a spreadsheet, for example. The CSV subdirectory contains a 24 | symlink "static", which points to the "static" subdirectory deployed by the pbench-web-server RPM. After you install the pbench-web-server RPM, you can just create a softlink to the javascript, like this: 25 | 26 | # yum install pbench-web-server 27 | # ln -sv static /var/www/html/static 28 | 29 | # server-side profiling 30 | 31 | Server-side profiling allows you to see activity across the entire Gluster volume for a specified number of periodic samples. It also allows you to see variation in stats between bricks, which can help you identify hotspots in your system where load is unevenly distributed. Results include: 32 | 33 | * per-volume MB/s read and written 34 | * per-brick MB/s read and written 35 | * per-volume per-FOP latency stats + call rates 36 | * per-brick per-FOP (File OPeration) latency stats + call rate 37 | 38 | It consists of: 39 | 40 | * gvp.sh: a bash script which runs the above command periodically for a number 41 | of samples, storing the results in a file. 42 | * extract_glvolprof.py: a python script that takes that output file 43 | and massages it into a form that can be used for visualization & analysis 44 | 45 | One component of this directory is an HTML file that can be viewed in a 46 | browser. The other is a bunch of CSV files containing the 47 | data. These files can also be used with a spreadsheet application if 48 | desired, to produce graphs that way 49 | 50 | Copy the scripts to some Gluster server in your cluster, (i.e. where you can run gluster volume profile command) and run the gvp.sh script. As an illustration, let's say we want to run it every 60 seconds and 10 iterations 51 | (10 minutes of operation) - in practice, you might want to 52 | do that periodically, perhaps in a cron job, in order to see the behavior 53 | of the cluster over time. 54 | 55 | \# ./gvp.sh [VOLNAME] 10 60 56 | 57 | Then run the extract script 58 | on that output file: 59 | 60 | \# python extract-glvolprof.py gvp.log 61 | 62 | The output (a bunch of CSV files and an HTML summary page) is placed in a subdirectory called gvp.log\_csvdir. 63 | 64 | To see the graphs, fire up a browser and point it to the URL that the extract script printed, pointing to gvp-graphs.html . 65 | 66 | # client-side profiling 67 | 68 | Client-side profiling allows you to see activity as close to the application as possible, at the top of the Gluster translator stack. This is particularly useful for identifying response time problems for the application related to Gluster activity. For example, Gluster replication causes a single application WRITE FOP to be transformed into multiple WRITE FOPs at the bricks within the volume where the file data resides. The response time for the application's WRITE request may be significantly different from the brick-level WRITE FOP latencies, because it incorporates the network response time and cannot complete before the brick-level WRITE FOPs complete. 69 | 70 | Copy the scripts to some directory on your client (i.e. where mountpoint is), and run the gvp-client.sh script. As an illustration, let's say we want to run it every 10 seconds and 12 iterations 71 | (roughly two minutes of operation) - in practice, you might want to 72 | do that periodically, perhaps in a cron job, in order to see the behavior 73 | of the cluster over time. 74 | 75 | \# ./gvp-client.sh [VOLNAME] [MOUNTPOINT] [SAMPLE-AMOUNT] [SAMPLE-DURATION-IN-SEC] 76 | \# ./gvp-client.sh vol1 /rhgs/client/vol1 12 10 77 | 78 | By default, the output file is called gvp-client-[Timestamp].log and saved in /var/tmp/. Then run the extract script 79 | on that output file: 80 | 81 | \# python extract-gl-client-prof.py /var/tmp/gvp-client-[Timestamp].log 82 | 83 | The output (a bunch of CSV files and an HTML summary page) is placed in 84 | a subdirectory in /var/tmp named similar to the supplied log file. 85 | 86 | To see the graphs, fire up a browser and point it to the URL that the extract 87 | script printed, pointing to gvp-client-graphs.html 88 | 89 | # implementation notes 90 | 91 | In order to take advantage of pbench javascript graphing, then column 1 in the .csv is always the timestamp in milliseconds when that sample took place. This can be disabled by defining the environment variable SKIP\_PBENCH\_GRAPHING. 92 | 93 | # appendix: detailed list of FOPs 94 | 95 | Here are all the file operation types that Gluster supports upstream as of November 2015. Looking for developers to correct descriptions here. The ones that are typically encountered are marked with the letter C: 96 | 97 | * ACCESS - ? 98 | * CREATE - C - create a file 99 | * DISCARD - support for trim? 100 | * ENTRYLK - lock a directory given its pathname? 101 | * FALLOCATE - allocate space for file without actually writing to it 102 | * FENTRYLK - lock a file given its handle 103 | * FGETXATTR - C - get named extended attribute value for a file (handle) 104 | * FINODELK - C - lock a file/directory for write/read 105 | * FLUSH - ensure all written data is persistently stored 106 | * FREMOVEXATTR - remove a named extended attribute from a file handle 107 | * FSETATTR - set value of metadata field (which ones?) for a file (handle) 108 | * FSETXATTR - C - set value of a named extended attribute for a file handle 109 | * FSTAT - get standard metadata about a file given its file handle 110 | * FSYNC - C - ensure all written data for a file is persistently stored 111 | * FSYNCDIR - ensure all directory entries in directory are persistently stored 112 | * FTRUNCATE - set file size to specified value, deallocating data beyond this point 113 | * FXATTROP - C - used by AFR replication? 114 | * GETXATTR - get value of named extended attribute 115 | * INODELK - lock a directory for write or for read 116 | * LINK - create a hard link 117 | * LK - lock? 118 | * LOOKUP - C - lookup file within directory 119 | * MKDIR - C - create directory 120 | * MKNOD - create device special file 121 | * OPEN - C - open a file 122 | * OPENDIR - C - open a directory (in preparation for READDIR) 123 | * RCHECKSUM - ? 124 | * READ - C - read data from a file 125 | * READDIR - C - read directory entries from a directory 126 | * READDIRP - C - read directory entries with standard metadata for each file (readdirplus) 127 | * READLINK - get the pathname of a file that a symlink is pointing to 128 | * RELEASE - C - let go of file handle (similar to close) 129 | * RELEASEDIR - let go of directory handle (similar to close) 130 | * REMOVEXATTR - remove a named extended attribute from a pathname? 131 | * RENAME - C - rename a file 132 | * RMDIR - C - remove a directory (assumes it is already empty) 133 | * SETATTR - set field in standard file metadata for pathname 134 | * SETXATTR - C - set named extended attribute value for file given pathname 135 | * STAT - C - get standard metadata for file given pathname 136 | * STATFS - get metadata for the filesystem 137 | * SYMLINK - create a softlink to specified pathname 138 | * TRUNCATE - truncate file at pathname to specified size 139 | * UNLINK - C - delete file 140 | * WRITE - C - write data to file 141 | * XATTROP - ? 142 | * ZEROFILL - write zeroes to the file in specified offset range 143 | -------------------------------------------------------------------------------- /gvp-README.html: -------------------------------------------------------------------------------- 1 | 2 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 96 | 142 | 143 | 144 |
145 |
146 |

Table of Contents

147 |
148 | 159 |
160 |
161 | 162 |
163 |

1 Introduction

164 |
165 |

166 | The extract-glvolprof.py program is meant to assist with visualizing the performance of 167 | a gluster volume, using the gluster volume profile command. 168 |

169 | 170 |

171 | It consists of three pieces: 172 |

173 | 174 |
    175 |
  • gvp.sh: a bash script which runs the above command periodically for a number 176 | of samples, storing the results in a file.
    • 177 | 178 |
  • extract_glvolprof.py: a python script that takes that output file 179 | and massages it into a form that can be used for visualization. One 180 | component of this is an HTMl file that can be viewed in a 181 | browser. The other is a bunch of CSV files containing the 182 | data. These files can also be used with a spreadsheet application if 183 | desired, to produce graphs that way.
    • 184 | 185 |
  • gvp-graph-javascript.tgz: a tarball containing some javascript 186 | libraries that are used by the HTML file above and provide common 187 | code to read CSV files and produce graphs using the nvd3 library.
    • 188 |
189 |
190 |
191 | 192 | 193 |
194 |

2 Installation and use

195 |
196 |
197 |

2.1 Simplest installation possible - somewhere on the cluster

198 |
199 |

200 | Copy the scripts to some directory in the cluster, unpack the tarball 201 | in the same directory. Run the gvp.sh script. As an illustration, 202 | let's say we want to run it every 10 seconds and 12 iterations 203 | (roughly two minutes of operation) - in practice, you might want to 204 | do that periodically, perhaps in a cron job, in order to see the behavior 205 | of the cluster over time: 206 |

207 |
208 | 209 | 
./gvp.sh 10 12
210 |
211 |
212 |

213 | By default, the output file is called gvp.log. Then run the extract script 214 | on that output file: 215 |

216 |
217 | 218 | 
python extract_glvolprof.py gvp.log
219 |
220 |
221 |

222 | The output (a bunch of CSV files and an HTML summary page) is placed in 223 | a subdirectory called gvp.log_csvdir. 224 |

225 |

226 | since we use pbench javascript graphing, then 227 | column 1 in the .csv is always the timestamp in milliseconds when 228 | that sample took place. This can be disabled with the environment variable 229 | SKIP_PBENCH_GRAPHING. 230 | 231 | the stat types are: 232 |

233 |
  • pct-lat - percentage latency consumed by this FOP (file operation)
    • 234 |
  • avg-lat - average latency (usec)
    • 235 |
  • min-lat - minimum latency (usec)
    • 236 |
  • max-lat - maximum latency (usec)
    • 237 |
  • call-rate - how many FOP requests have been processed per second
    • 238 |

    239 | for each category, there are several kinds of .csv files produced. 240 | For each FOP + stat type, we show per-brick results and results across all bricks. 241 | For each stat type, across all bricks, we show FOP stats together in vol*allfop.csv files. 242 | We also show read/write traffic by brick in MBps*bricks.csv files. 243 | Finally we show read/write traffic for the entire volume in 244 | MBps*volume.csv files. 245 |

    246 |

    247 | To see the graphs, fire up a browser and point it to the URL that the extract 248 | script printed: 249 |

    250 |
    251 | 252 | 
    firefox gvp.log_csvdir/gvp-graphs.html
253 |
    254 |
    255 |

    256 | BTW, not everything works: e.g. the "Save as Image" button does not. Note 257 | also that the layout is crucial: the CSV subdirectory contains a 258 | symlink static, which points to the static subdirectory in the 259 | main directory (which is where the javascript tarball was 260 | unpacked). If you change that structure, then the javascript files may 261 | not be found: no graphs! 262 |

    263 |
    264 |
    265 | 266 |
    267 |

    2.2 Simplest installation possible - on your own computer

    268 |
    269 |

    270 | Alternatively, install everything as above somewhere on your own computer. 271 | Copy only the gvp.sh script to the cluster and run it there, then copy 272 | the gvp.log file back to your computer and process it locally as above. 273 |

    274 |
    275 |
    276 | 277 |
    278 |

    2.3 More general installation (requires HTTP server)

    279 |
    280 |

    281 | For more general access, you might want to set up an HTTP server, either 282 | in the cluster itself or on some other accessible machine. 283 |

    284 | 285 |

    286 | The idea is to run the gvp.sh script in the cluster and copy the 287 | gvp.log file (probably renaming it using a timestamp to keep the 288 | name unique) to the HTTP server, under the document root directory. 289 | The extraction script can then be run on it to produce the CSV 290 | subdirectory under the server's document root: 291 |

    292 | 293 |
    294 | 295 | 
    python extract_glvolprof.py /var/www/html/gvp/gvp.log.TIMESTAMP
296 |
    297 |
    298 | 299 |

    300 | You can then view the results from a browser running elsewhere using 301 | a URL like 302 |

    303 | 
    304 | http://SERVER.DOMAIN/gvp/gvp.log.TIMESTAMP_csvdir/gvp-graphs.html
305 |
    306 |
    307 |
    308 | 309 |
    310 |

    2.4 Using ElasticSearch and Grafana - future direction

    311 |
    312 |

    313 | It's also possible to index the CSV files into ElasticSearch and use 314 | Grafana for the visualization. That's probably the most flexible solution, 315 | but it would need some additional work, not included here. 316 |

    317 |
    318 |
    319 |
    320 |
    321 |
    322 |

    Created: 2015-10-26 Mon 19:38

    323 |

    Validate

    324 |
    325 | 326 | 327 | -------------------------------------------------------------------------------- /LICENSE: -------------------------------------------------------------------------------- 1 | GNU GENERAL PUBLIC LICENSE 2 | Version 2, June 1991 3 | 4 | Copyright (C) 1989, 1991 Free Software Foundation, Inc., 5 | 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA 6 | Everyone is permitted to copy and distribute verbatim copies 7 | of this license document, but changing it is not allowed. 8 | 9 | Preamble 10 | 11 | The licenses for most software are designed to take away your 12 | freedom to share and change it. By contrast, the GNU General Public 13 | License is intended to guarantee your freedom to share and change free 14 | software--to make sure the software is free for all its users. This 15 | General Public License applies to most of the Free Software 16 | Foundation's software and to any other program whose authors commit to 17 | using it. (Some other Free Software Foundation software is covered by 18 | the GNU Lesser General Public License instead.) You can apply it to 19 | your programs, too. 20 | 21 | When we speak of free software, we are referring to freedom, not 22 | price. Our General Public Licenses are designed to make sure that you 23 | have the freedom to distribute copies of free software (and charge for 24 | this service if you wish), that you receive source code or can get it 25 | if you want it, that you can change the software or use pieces of it 26 | in new free programs; and that you know you can do these things. 27 | 28 | To protect your rights, we need to make restrictions that forbid 29 | anyone to deny you these rights or to ask you to surrender the rights. 30 | These restrictions translate to certain responsibilities for you if you 31 | distribute copies of the software, or if you modify it. 32 | 33 | For example, if you distribute copies of such a program, whether 34 | gratis or for a fee, you must give the recipients all the rights that 35 | you have. You must make sure that they, too, receive or can get the 36 | source code. And you must show them these terms so they know their 37 | rights. 38 | 39 | We protect your rights with two steps: (1) copyright the software, and 40 | (2) offer you this license which gives you legal permission to copy, 41 | distribute and/or modify the software. 42 | 43 | Also, for each author's protection and ours, we want to make certain 44 | that everyone understands that there is no warranty for this free 45 | software. If the software is modified by someone else and passed on, we 46 | want its recipients to know that what they have is not the original, so 47 | that any problems introduced by others will not reflect on the original 48 | authors' reputations. 49 | 50 | Finally, any free program is threatened constantly by software 51 | patents. We wish to avoid the danger that redistributors of a free 52 | program will individually obtain patent licenses, in effect making the 53 | program proprietary. To prevent this, we have made it clear that any 54 | patent must be licensed for everyone's free use or not licensed at all. 55 | 56 | The precise terms and conditions for copying, distribution and 57 | modification follow. 58 | 59 | GNU GENERAL PUBLIC LICENSE 60 | TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION 61 | 62 | 0. This License applies to any program or other work which contains 63 | a notice placed by the copyright holder saying it may be distributed 64 | under the terms of this General Public License. The "Program", below, 65 | refers to any such program or work, and a "work based on the Program" 66 | means either the Program or any derivative work under copyright law: 67 | that is to say, a work containing the Program or a portion of it, 68 | either verbatim or with modifications and/or translated into another 69 | language. (Hereinafter, translation is included without limitation in 70 | the term "modification".) Each licensee is addressed as "you". 71 | 72 | Activities other than copying, distribution and modification are not 73 | covered by this License; they are outside its scope. The act of 74 | running the Program is not restricted, and the output from the Program 75 | is covered only if its contents constitute a work based on the 76 | Program (independent of having been made by running the Program). 77 | Whether that is true depends on what the Program does. 78 | 79 | 1. You may copy and distribute verbatim copies of the Program's 80 | source code as you receive it, in any medium, provided that you 81 | conspicuously and appropriately publish on each copy an appropriate 82 | copyright notice and disclaimer of warranty; keep intact all the 83 | notices that refer to this License and to the absence of any warranty; 84 | and give any other recipients of the Program a copy of this License 85 | along with the Program. 86 | 87 | You may charge a fee for the physical act of transferring a copy, and 88 | you may at your option offer warranty protection in exchange for a fee. 89 | 90 | 2. You may modify your copy or copies of the Program or any portion 91 | of it, thus forming a work based on the Program, and copy and 92 | distribute such modifications or work under the terms of Section 1 93 | above, provided that you also meet all of these conditions: 94 | 95 | a) You must cause the modified files to carry prominent notices 96 | stating that you changed the files and the date of any change. 97 | 98 | b) You must cause any work that you distribute or publish, that in 99 | whole or in part contains or is derived from the Program or any 100 | part thereof, to be licensed as a whole at no charge to all third 101 | parties under the terms of this License. 102 | 103 | c) If the modified program normally reads commands interactively 104 | when run, you must cause it, when started running for such 105 | interactive use in the most ordinary way, to print or display an 106 | announcement including an appropriate copyright notice and a 107 | notice that there is no warranty (or else, saying that you provide 108 | a warranty) and that users may redistribute the program under 109 | these conditions, and telling the user how to view a copy of this 110 | License. (Exception: if the Program itself is interactive but 111 | does not normally print such an announcement, your work based on 112 | the Program is not required to print an announcement.) 113 | 114 | These requirements apply to the modified work as a whole. If 115 | identifiable sections of that work are not derived from the Program, 116 | and can be reasonably considered independent and separate works in 117 | themselves, then this License, and its terms, do not apply to those 118 | sections when you distribute them as separate works. But when you 119 | distribute the same sections as part of a whole which is a work based 120 | on the Program, the distribution of the whole must be on the terms of 121 | this License, whose permissions for other licensees extend to the 122 | entire whole, and thus to each and every part regardless of who wrote it. 123 | 124 | Thus, it is not the intent of this section to claim rights or contest 125 | your rights to work written entirely by you; rather, the intent is to 126 | exercise the right to control the distribution of derivative or 127 | collective works based on the Program. 128 | 129 | In addition, mere aggregation of another work not based on the Program 130 | with the Program (or with a work based on the Program) on a volume of 131 | a storage or distribution medium does not bring the other work under 132 | the scope of this License. 133 | 134 | 3. You may copy and distribute the Program (or a work based on it, 135 | under Section 2) in object code or executable form under the terms of 136 | Sections 1 and 2 above provided that you also do one of the following: 137 | 138 | a) Accompany it with the complete corresponding machine-readable 139 | source code, which must be distributed under the terms of Sections 140 | 1 and 2 above on a medium customarily used for software interchange; or, 141 | 142 | b) Accompany it with a written offer, valid for at least three 143 | years, to give any third party, for a charge no more than your 144 | cost of physically performing source distribution, a complete 145 | machine-readable copy of the corresponding source code, to be 146 | distributed under the terms of Sections 1 and 2 above on a medium 147 | customarily used for software interchange; or, 148 | 149 | c) Accompany it with the information you received as to the offer 150 | to distribute corresponding source code. (This alternative is 151 | allowed only for noncommercial distribution and only if you 152 | received the program in object code or executable form with such 153 | an offer, in accord with Subsection b above.) 154 | 155 | The source code for a work means the preferred form of the work for 156 | making modifications to it. For an executable work, complete source 157 | code means all the source code for all modules it contains, plus any 158 | associated interface definition files, plus the scripts used to 159 | control compilation and installation of the executable. However, as a 160 | special exception, the source code distributed need not include 161 | anything that is normally distributed (in either source or binary 162 | form) with the major components (compiler, kernel, and so on) of the 163 | operating system on which the executable runs, unless that component 164 | itself accompanies the executable. 165 | 166 | If distribution of executable or object code is made by offering 167 | access to copy from a designated place, then offering equivalent 168 | access to copy the source code from the same place counts as 169 | distribution of the source code, even though third parties are not 170 | compelled to copy the source along with the object code. 171 | 172 | 4. You may not copy, modify, sublicense, or distribute the Program 173 | except as expressly provided under this License. Any attempt 174 | otherwise to copy, modify, sublicense or distribute the Program is 175 | void, and will automatically terminate your rights under this License. 176 | However, parties who have received copies, or rights, from you under 177 | this License will not have their licenses terminated so long as such 178 | parties remain in full compliance. 179 | 180 | 5. You are not required to accept this License, since you have not 181 | signed it. However, nothing else grants you permission to modify or 182 | distribute the Program or its derivative works. These actions are 183 | prohibited by law if you do not accept this License. Therefore, by 184 | modifying or distributing the Program (or any work based on the 185 | Program), you indicate your acceptance of this License to do so, and 186 | all its terms and conditions for copying, distributing or modifying 187 | the Program or works based on it. 188 | 189 | 6. Each time you redistribute the Program (or any work based on the 190 | Program), the recipient automatically receives a license from the 191 | original licensor to copy, distribute or modify the Program subject to 192 | these terms and conditions. You may not impose any further 193 | restrictions on the recipients' exercise of the rights granted herein. 194 | You are not responsible for enforcing compliance by third parties to 195 | this License. 196 | 197 | 7. If, as a consequence of a court judgment or allegation of patent 198 | infringement or for any other reason (not limited to patent issues), 199 | conditions are imposed on you (whether by court order, agreement or 200 | otherwise) that contradict the conditions of this License, they do not 201 | excuse you from the conditions of this License. If you cannot 202 | distribute so as to satisfy simultaneously your obligations under this 203 | License and any other pertinent obligations, then as a consequence you 204 | may not distribute the Program at all. For example, if a patent 205 | license would not permit royalty-free redistribution of the Program by 206 | all those who receive copies directly or indirectly through you, then 207 | the only way you could satisfy both it and this License would be to 208 | refrain entirely from distribution of the Program. 209 | 210 | If any portion of this section is held invalid or unenforceable under 211 | any particular circumstance, the balance of the section is intended to 212 | apply and the section as a whole is intended to apply in other 213 | circumstances. 214 | 215 | It is not the purpose of this section to induce you to infringe any 216 | patents or other property right claims or to contest validity of any 217 | such claims; this section has the sole purpose of protecting the 218 | integrity of the free software distribution system, which is 219 | implemented by public license practices. Many people have made 220 | generous contributions to the wide range of software distributed 221 | through that system in reliance on consistent application of that 222 | system; it is up to the author/donor to decide if he or she is willing 223 | to distribute software through any other system and a licensee cannot 224 | impose that choice. 225 | 226 | This section is intended to make thoroughly clear what is believed to 227 | be a consequence of the rest of this License. 228 | 229 | 8. If the distribution and/or use of the Program is restricted in 230 | certain countries either by patents or by copyrighted interfaces, the 231 | original copyright holder who places the Program under this License 232 | may add an explicit geographical distribution limitation excluding 233 | those countries, so that distribution is permitted only in or among 234 | countries not thus excluded. In such case, this License incorporates 235 | the limitation as if written in the body of this License. 236 | 237 | 9. The Free Software Foundation may publish revised and/or new versions 238 | of the General Public License from time to time. Such new versions will 239 | be similar in spirit to the present version, but may differ in detail to 240 | address new problems or concerns. 241 | 242 | Each version is given a distinguishing version number. If the Program 243 | specifies a version number of this License which applies to it and "any 244 | later version", you have the option of following the terms and conditions 245 | either of that version or of any later version published by the Free 246 | Software Foundation. If the Program does not specify a version number of 247 | this License, you may choose any version ever published by the Free Software 248 | Foundation. 249 | 250 | 10. If you wish to incorporate parts of the Program into other free 251 | programs whose distribution conditions are different, write to the author 252 | to ask for permission. For software which is copyrighted by the Free 253 | Software Foundation, write to the Free Software Foundation; we sometimes 254 | make exceptions for this. Our decision will be guided by the two goals 255 | of preserving the free status of all derivatives of our free software and 256 | of promoting the sharing and reuse of software generally. 257 | 258 | NO WARRANTY 259 | 260 | 11. BECAUSE THE PROGRAM IS LICENSED FREE OF CHARGE, THERE IS NO WARRANTY 261 | FOR THE PROGRAM, TO THE EXTENT PERMITTED BY APPLICABLE LAW. EXCEPT WHEN 262 | OTHERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR OTHER PARTIES 263 | PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED 264 | OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF 265 | MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE RISK AS 266 | TO THE QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU. SHOULD THE 267 | PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL NECESSARY SERVICING, 268 | REPAIR OR CORRECTION. 269 | 270 | 12. IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING 271 | WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MAY MODIFY AND/OR 272 | REDISTRIBUTE THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, 273 | INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING 274 | OUT OF THE USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED 275 | TO LOSS OF DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY 276 | YOU OR THIRD PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER 277 | PROGRAMS), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE 278 | POSSIBILITY OF SUCH DAMAGES. 279 | 280 | END OF TERMS AND CONDITIONS 281 | 282 | How to Apply These Terms to Your New Programs 283 | 284 | If you develop a new program, and you want it to be of the greatest 285 | possible use to the public, the best way to achieve this is to make it 286 | free software which everyone can redistribute and change under these terms. 287 | 288 | To do so, attach the following notices to the program. It is safest 289 | to attach them to the start of each source file to most effectively 290 | convey the exclusion of warranty; and each file should have at least 291 | the "copyright" line and a pointer to where the full notice is found. 292 | 293 | {description} 294 | Copyright (C) {year} {fullname} 295 | 296 | This program is free software; you can redistribute it and/or modify 297 | it under the terms of the GNU General Public License as published by 298 | the Free Software Foundation; either version 2 of the License, or 299 | (at your option) any later version. 300 | 301 | This program is distributed in the hope that it will be useful, 302 | but WITHOUT ANY WARRANTY; without even the implied warranty of 303 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 304 | GNU General Public License for more details. 305 | 306 | You should have received a copy of the GNU General Public License along 307 | with this program; if not, write to the Free Software Foundation, Inc., 308 | 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA. 309 | 310 | Also add information on how to contact you by electronic and paper mail. 311 | 312 | If the program is interactive, make it output a short notice like this 313 | when it starts in an interactive mode: 314 | 315 | Gnomovision version 69, Copyright (C) year name of author 316 | Gnomovision comes with ABSOLUTELY NO WARRANTY; for details type `show w'. 317 | This is free software, and you are welcome to redistribute it 318 | under certain conditions; type `show c' for details. 319 | 320 | The hypothetical commands `show w' and `show c' should show the appropriate 321 | parts of the General Public License. Of course, the commands you use may 322 | be called something other than `show w' and `show c'; they could even be 323 | mouse-clicks or menu items--whatever suits your program. 324 | 325 | You should also get your employer (if you work as a programmer) or your 326 | school, if any, to sign a "copyright disclaimer" for the program, if 327 | necessary. Here is a sample; alter the names: 328 | 329 | Yoyodyne, Inc., hereby disclaims all copyright interest in the program 330 | `Gnomovision' (which makes passes at compilers) written by James Hacker. 331 | 332 | {signature of Ty Coon}, 1 April 1989 333 | Ty Coon, President of Vice 334 | 335 | This General Public License does not permit incorporating your program into 336 | proprietary programs. If your program is a subroutine library, you may 337 | consider it more useful to permit linking proprietary applications with the 338 | library. If this is what you want to do, use the GNU Lesser General 339 | Public License instead of this License. 340 | 341 | -------------------------------------------------------------------------------- /extract-gl-client-prof.py: -------------------------------------------------------------------------------- 1 | #!/usr/bin/python 2 | # -*- coding: utf-8 -*- 3 | 4 | # 5 | # extract-gl-client-prof.py 6 | # written by Ben England 2015 7 | # copyright is GNU GPL V3, for details read: 8 | # https://tldrlegal.com/license/gnu-general-public-license-v3-%28gpl-3%29#fulltext 9 | # 10 | # Note: this tool uses a snapshot of javascript code from this project: 11 | # https://github.com/distributed-system-analysis/pbench 12 | # but we do not support any use of this software outside of the graphing 13 | # of the data generated below. 14 | # 15 | # script to read gluster client-side output retrieved every N seconds 16 | # and generate operation rate graph from it 17 | # 18 | # NOTE: the tool creates a subdirectory just for each run of this analysis tool. 19 | # the directory name is just the name of the log file 20 | # with the suffix '_csvdir' 21 | # 22 | # to install: 23 | # - extract javascript code from this tarball 24 | # https://s3.amazonaws.com/ben.england/gvp-graph-javascript.tgz 25 | # - if the directory containing your gluster volume output log is different, create a 26 | # 'static' symlink pointing to the static/ subdirectory you just extracted 27 | # in the subdirectories where .csv and .html files live, you will see a 28 | # 'static' softlink pointing to this symlink. 29 | # 30 | # input: 31 | # this script expects input data to look like what this script produces: 32 | # 33 | # https://raw.githubusercontent.com/bengland2/parallel-libgfapi/master/gvp-client.sh 34 | # 35 | # record 1 contains the user-specified sample count and interval 36 | # used by gvp-client.sh. 37 | # record 2 is a timestamp generated by gluster in format like: 38 | # Wed Oct 21 22:50:28 UTC 2015 39 | # subsequent "gluster volume profile your-volume info" outputs are 40 | # concatenated to the profile log. 41 | # Each profile sample is assumed to happen approximately N seconds after 42 | # the preceding sample, where N is the gvp.sh sampling interval. 43 | # seconds. The first sample happens N seconds after the timestamp. 44 | # 45 | # output: 46 | # 47 | # when we're all done reading in data, 48 | # we then print it out in a format suitable for spreadsheet-based graphing 49 | # 50 | # since we use pbench javascript graphing, then 51 | # column 1 in the .csv is always the timestamp in milliseconds when 52 | # that sample took place. This can be disabled with the environment variable 53 | # SKIP_PBENCH_GRAPHING. 54 | # 55 | # the stat types are: 56 | # - pct-lat - percentage latency consumed by this FOP (file operation) 57 | # - avg-lat - average latency (usec) 58 | # - min-lat - minimum latency (usec) 59 | # - max-lat - maximum latency (usec) 60 | # - call-rate - how many FOP requests have been processed per second 61 | # for each category: 62 | # - for each stat type, show stat by FOP 63 | # 64 | # internals: 65 | # 66 | # the "intervals" array, indexed by interval number, stores results over time 67 | # within each array element, we have IntervalProfile objects containing 68 | # bytes read/written and a dictionary indexed by FOP name 69 | # containing FopProfile instances to represent the per-FOP records 70 | # in "gluster volume profile" output. 71 | # the per-FOP dictionary is indexed by FOP name 72 | # 73 | 74 | import sys 75 | import os 76 | from os.path import join 77 | import re 78 | import time 79 | import shutil 80 | import collections 81 | 82 | # fields in gluster volume profile output 83 | 84 | stat_names = ['pct-lat', 'avg-lat', 'min-lat', 'max-lat', 'call-rate'] 85 | directions = ['MBps-read', 'MBps-written'] 86 | min_lat_infinity = 1.0e24 87 | 88 | # this environment variable lets you graph .csv files using pbench 89 | 90 | pbench_graphs = True 91 | if os.getenv('SKIP_PBENCH_GRAPHING'): pbench_graphs = False 92 | 93 | # this is the list of graphs that will be produced 94 | 95 | graph_csvs = [ 96 | ('MBps-written', 'MB/sec written to Gluster volume'), 97 | ('MBps-read', 'MB/sec read from Gluster volume'), 98 | ('call-rate', 'FOP call rates'), 99 | ('pct-lat', 'percentage latency by FOP') 100 | ] 101 | 102 | # all gvp.sh-generated profiles are expected to have these parameters 103 | # we define them here to have global scope, and they are only changed 104 | # by the input parser 105 | 106 | start_time = None 107 | expected_duration = None 108 | expected_sample_count = None 109 | sorted_fop_names = None 110 | intervals = None 111 | 112 | # this class stores per-fop statistics from gluster client profile output 113 | # to compute stats for %latency and average latency across a set of bricks, 114 | # we have to compute averages weighted by FOP calls 115 | # We do this in two steps: 116 | # - loop over set of instances and compute weighted sum (not average) 117 | # - after loop, normalize using total calls 118 | 119 | 120 | class FopProfile: 121 | 122 | def __init__(self, avg_lat, min_lat, max_lat, calls): 123 | self.avg_lat = avg_lat 124 | self.min_lat = min_lat 125 | self.max_lat = max_lat 126 | self.calls = calls 127 | self.pct_lat = 0.0 # will compute later 128 | 129 | def __str__(self): 130 | return '%6.2f, %8.0f, %8.0f, %8.0f, %d' % ( 131 | self.pct_lat, self.avg_lat, self.min_lat, self.max_lat, self.calls) 132 | 133 | # append a single field to .csv record based on statistic type 134 | # use "-6.2f" instead of "%6.2f" so there are no leading spaces in record, 135 | # otherwise spreadsheet inserts colums at col. B 136 | 137 | def field2str(self, stat, duration): 138 | if stat == stat_names[0]: 139 | return '%-6.2f' % self.pct_lat 140 | elif stat == stat_names[1]: 141 | return '%8.0f' % self.avg_lat 142 | elif stat == stat_names[2]: 143 | if self.min_lat == min_lat_infinity: 144 | return '' # don't confuse spreadsheet/user 145 | else: 146 | return '%8.0f' % self.min_lat 147 | elif stat == stat_names[3]: 148 | if self.max_lat == 0: 149 | return '' 150 | else: 151 | return '%8.0f' % self.max_lat 152 | elif stat == stat_names[4]: 153 | call_rate = self.calls / float(duration) 154 | return '%10.3f' % call_rate 155 | 156 | # accumulate weighted sum of component profiles, will normalize them later 157 | 158 | def accumulate(self, addend): 159 | self.avg_lat += (addend.avg_lat * addend.calls) 160 | self.calls += addend.calls 161 | if addend.calls > 0: 162 | self.max_lat = max(self.max_lat, addend.max_lat) 163 | self.min_lat = min(self.min_lat, addend.min_lat) 164 | 165 | # normalize weighted sum to get averages 166 | 167 | def normalize_sum(self): 168 | try: 169 | # totals will become averages 170 | self.avg_lat /= self.calls 171 | except ZeroDivisionError: # if no samples, set these stats to zero 172 | self.pct_lat = 0.0 173 | self.avg_lat = 0.0 174 | 175 | # compute % latency for this FOP given total latency of all FOPs 176 | 177 | def get_pct_lat(self, total_lat): 178 | try: 179 | self.pct_lat = 100.0 * (self.avg_lat * self.calls) / total_lat 180 | except ZeroDivisionError: # if no samples, set these stats to zero 181 | self.pct_lat = 0.0 182 | 183 | 184 | class ProfileInterval: 185 | 186 | def __init__(self): 187 | self.bytes_read = None 188 | self.bytes_written = None 189 | self.duration = None 190 | self.fop_profiles = {} 191 | 192 | def __str__(self): 193 | return '%d, %d, %s, %s'%( 194 | self.bytes_read, self.bytes_written, 195 | str(self.duration), [ str(f) + ' : ' + str(self.fop_profiles[f]) for f in self.fop_profiles ]) 196 | 197 | 198 | # if there is an error parsing the input... 199 | 200 | def usage(msg): 201 | print('ERROR: %s' % msg) 202 | print('usage: extract-gl-client-prof.py your-gluster-client-profile.log') 203 | sys.exit(1) 204 | 205 | 206 | # segregate .csv files into a separate output directory 207 | # with pathname derived from the input log file with _csvdir suffix 208 | 209 | def make_out_dir(path): 210 | dir_path = path + '_csvdir' 211 | try: 212 | if os.path.exists(dir_path): 213 | shutil.rmtree(dir_path) 214 | os.mkdir(dir_path) 215 | except IOError: 216 | usage('could not (re-)create directory ' + dir_path) 217 | return dir_path 218 | 219 | 220 | # convert gvp-client.sh client profile output 221 | # into a time series of per-fop results. 222 | 223 | def parse_input(input_pathname): 224 | global start_time 225 | global expected_sample_interval 226 | global expected_sample_count 227 | global sorted_fop_names 228 | global intervals 229 | 230 | try: 231 | with open(input_pathname, 'r') as file_handle: 232 | lines = [ l.strip() for l in file_handle.readlines() ] 233 | except IOError: 234 | usage('could not read ' + input_pathname) 235 | tokens = lines[0].split() 236 | expected_sample_count = int(tokens[0]) 237 | expected_sample_interval = int(tokens[1]) 238 | start_time = time.mktime( 239 | time.strptime( 240 | lines[1], '%a %b %d %H:%M:%S %Z %Y')) * 1000 241 | print('collection started at %s' % lines[1]) 242 | print('sampling interval is %d seconds' % expected_sample_interval) 243 | print('expected sample count is %d samples' % expected_sample_count) 244 | 245 | # parse the file and record each cell of output in a way that lets you 246 | # aggregate across bricks later 247 | 248 | found_cumulative_output = False 249 | found_interval_output = False 250 | all_caps_name = re.compile('^[A-Z]{3,15}') 251 | fop_names = set() 252 | last_intvl = -2 253 | intvl = -1 254 | per_op_table = {} 255 | sample = -1 256 | intervals = [] 257 | for ln in lines[2:]: 258 | tokens = ln.split() 259 | 260 | if ln.__contains__('Interval') and ln.__contains__('stats'): 261 | 262 | interval_number = int(tokens[2]) 263 | assert intvl == last_intvl + 1 264 | last_intvl = intvl 265 | intvl += 1 266 | intvl_profile = ProfileInterval() 267 | intervals.append(intvl_profile) 268 | found_interval_output = True 269 | 270 | elif ln.__contains__('Cumulative Stats'): 271 | 272 | found_cumulative_output = True 273 | 274 | elif ln.__contains__('Duration :'): 275 | 276 | # we are at end of output for this brick and interval 277 | 278 | assert found_cumulative_output ^ found_interval_output 279 | duration = int(tokens[2]) 280 | diff_from_expected = abs(duration - expected_sample_interval) 281 | if found_interval_output: 282 | if diff_from_expected > 1: 283 | print(('WARNING: in sample %d the sample ' + 284 | 'interval %d deviates from expected value %d') % 285 | (sample, duration, expected_sample_interval)) 286 | fops_in_interval = intervals[intvl] 287 | fops_in_interval.duration = duration 288 | 289 | elif ln.__contains__('BytesRead'): 290 | 291 | if found_interval_output: 292 | intvl_profile = intervals[intvl] 293 | intvl_profile.bytes_read = int(tokens[2]) 294 | 295 | elif ln.__contains__('BytesWritten'): 296 | 297 | if found_interval_output: 298 | intvl_profile = intervals[intvl] 299 | intvl_profile.bytes_written = int(tokens[2]) 300 | 301 | elif ln.__contains__('Cumulative stats'): 302 | 303 | # this is the end of this sample 304 | 305 | found_interval_output = False 306 | found_cumulative_output = True 307 | 308 | elif ln.__contains__('Current open fd'): 309 | 310 | found_cumulative_output = False 311 | 312 | elif found_interval_output and all_caps_name.match(ln): 313 | 314 | # we found a record we're interested in, 315 | # accumulate table of data for each gluster function 316 | 317 | sample += 1 318 | intvl_profile = intervals[intvl] 319 | fop_name = tokens[0] 320 | fop_names.add(fop_name) 321 | new_fop_profile = FopProfile( 322 | float(tokens[2]), float(tokens[4]), float(tokens[6]), 323 | float(tokens[1])) 324 | try: 325 | fop_stats = intvl_profile.fop_profiles[fop_name] 326 | raise Exception('did not expect fop already defined: %s' % 327 | str(intvl_profile)) 328 | except KeyError: 329 | intvl_profile.fop_profiles[fop_name] = new_fop_profile 330 | sorted_fop_names = sorted(fop_names) 331 | 332 | 333 | # generate timestamp_ms column for pbench 334 | # given starting time of collection, sampling interval and sample number 335 | 336 | def gen_timestamp_ms(sample_index): 337 | return start_time + ((expected_sample_interval * sample_index) * 1000) 338 | 339 | 340 | # generate denominator for call rate computation based on duration type 341 | 342 | def get_interval(interval_index, duration_type = 'interval'): 343 | if duration_type == 'cumulative': 344 | return interval_index * float(expected_sample_interval) 345 | else: 346 | return float(expected_sample_interval) 347 | 348 | # display bytes read and bytes written 349 | # normalize to MB/s with 3 decimal places so 1 KB/s/brick will show 350 | 351 | def gen_output_bytes(out_dir_path): 352 | bytes_per_MB = 1000000.0 353 | for direction in directions: 354 | # when we support cumulative data, then we can name files this way 355 | #direction_filename = duration_type + '_' + direction + '.csv' 356 | direction_filename = direction + '.csv' 357 | direction_pathname = join(out_dir_path, direction_filename) 358 | with open(direction_pathname, 'w') as transfer_fh: 359 | if pbench_graphs: 360 | transfer_fh.write('timestamp_ms, ') 361 | transfer_fh.write('MB/s\n') 362 | for j in range(0, len(intervals)): 363 | if pbench_graphs: 364 | transfer_fh.write('%d, ' % gen_timestamp_ms(j)) 365 | rate_interval = get_interval(j) 366 | interval_profile = intervals[j] 367 | if direction.__contains__('read'): 368 | transfer = interval_profile.bytes_read 369 | else: 370 | transfer = interval_profile.bytes_written 371 | transfer_fh.write('%-8.3f\n' % 372 | ((transfer/rate_interval)/bytes_per_MB)) 373 | 374 | # display per-FOP (file operation) stats, 375 | 376 | def gen_per_fop_stats(out_dir_path, stat, duration_type='interval'): 377 | per_fop_filename = stat + '.csv' 378 | per_fop_path = join(out_dir_path, per_fop_filename) 379 | with open(per_fop_path, 'a') as fop_fh: 380 | hdr = '' 381 | if pbench_graphs: 382 | hdr += 'timestamp_ms, ' 383 | hdr += ','.join(sorted_fop_names) 384 | hdr += '\n' 385 | fop_fh.write(hdr) 386 | for i in range(0, len(intervals)): 387 | interval_profile = intervals[i] 388 | fops_in_interval = interval_profile.fop_profiles 389 | all_fop_profile = FopProfile(0, 0, 0, 0) 390 | for fop in sorted_fop_names: 391 | fop_stats = fops_in_interval[fop] 392 | all_fop_profile.accumulate(fop_stats) 393 | all_fop_profile.normalize_sum() 394 | #print('intvl: %d' % i) 395 | #print('ALL FOPs: %s' % all_fop_profile) 396 | if pbench_graphs: 397 | fop_fh.write('%d, ' % gen_timestamp_ms(i)) 398 | columns = [] 399 | for fop in sorted_fop_names: 400 | fop_stats = fops_in_interval[fop] 401 | fop_stats.get_pct_lat( 402 | all_fop_profile.avg_lat * all_fop_profile.calls) 403 | try: 404 | fop_stats = fops_in_interval[fop] 405 | except KeyError: 406 | fops_in_interval[fop] = fop_stats 407 | columns.append( 408 | fop_stats.field2str( 409 | stat, interval_profile.duration)) 410 | fop_fh.write(','.join(columns) + '\n') 411 | 412 | # generate graphs in 413 | # generate output files in separate directory from 414 | # data structure returned by parse_input 415 | 416 | next_graph_template=''' 417 |
    418 |

    %s 419 | 420 |
    421 |

    422 | 423 | 424 | 427 |
    428 | ''' 429 | 430 | def output_next_graph(graph_fh, gr_index): 431 | (csv_filename, graph_description) = graph_csvs[gr_index] 432 | gr_index += 1 # graph numbers start at 1 433 | graph_fh.write( next_graph_template % ( 434 | graph_description, gr_index, gr_index, csv_filename)) 435 | 436 | # static content of HTML file 437 | 438 | header=''' 439 | 440 | 441 | 442 | 443 | 444 | 445 | 446 | 447 | 448 | 449 | 450 | 451 | 452 | 453 |

    summary profile of application activity on one client

    454 | ''' 455 | 456 | trailer=''' 457 | 458 | 459 | ''' 460 | 461 | 462 | # generate graphs using header, trailer and graph template 463 | 464 | def gen_graphs(out_dir_path): 465 | graph_path = join(out_dir_path, 'gvp-client-graphs.html') 466 | with open(graph_path, 'w') as graph_fh: 467 | graph_fh.write(header) 468 | for j in range(0, len(graph_csvs)): 469 | output_next_graph(graph_fh, j) 470 | graph_fh.write(trailer) 471 | return graph_path 472 | 473 | 474 | # make link to where javascript etc lives in unpacked tarball 475 | # ASSUMPTION is that output directory is a subdirectory of where this script 476 | # lives (not a sub-subdirectory). Sorry but that's the only way to generate a 477 | # softlink that works when we copy the csvdir to a different location. 478 | 479 | def gen_static_softlink(out_dir_path): 480 | saved_cwd = os.getcwd() 481 | static_dir = join(saved_cwd, 'static') 482 | if not os.path.exists(static_dir): 483 | print('ERROR: sorry, the javascript directory "static" ' + 484 | 'needs to be in same directory as this script, trying anyway...') 485 | os.chdir(out_dir_path) 486 | os.symlink(join('..', 'static'), 'static') 487 | os.chdir(saved_cwd) 488 | 489 | # generate everything needed to view the graphs 490 | 491 | def generate_output(out_dir_path): 492 | 493 | gen_output_bytes(out_dir_path) 494 | for s in stat_names: 495 | gen_per_fop_stats(out_dir_path, s) 496 | graph_path = gen_graphs(out_dir_path) 497 | gen_static_softlink(out_dir_path) 498 | 499 | sys.stdout.write('Gluster FOP types seen: ') 500 | for fop_name in sorted_fop_names: 501 | sys.stdout.write(' ' + fop_name) 502 | sys.stdout.write('\n') 503 | print('created Gluster statistics files in directory %s' % out_dir_path) 504 | if not os.path.isabs(graph_path): 505 | graph_path = join(os.getcwd(), graph_path) 506 | print('graphs now available at browser URL file://%s' % graph_path) 507 | 508 | 509 | # the main program is kept in a subroutine so that it can run on Windows. 510 | 511 | def main(): 512 | if len(sys.argv) < 2: 513 | usage('missing gluster volume profile output log filename parameter' 514 | ) 515 | fn = sys.argv[1] 516 | parse_input(fn) 517 | outdir = make_out_dir(fn) 518 | generate_output(outdir) 519 | 520 | main() 521 | -------------------------------------------------------------------------------- /extract-glvolprof.py: -------------------------------------------------------------------------------- 1 | #!/usr/bin/python 2 | # -*- coding: utf-8 -*- 3 | 4 | # 5 | # extract-glvolprof.py 6 | # written by Ben England 2015 7 | # copyright is GNU GPL V3, for details read: 8 | # https://tldrlegal.com/license/gnu-general-public-license-v3-%28gpl-3%29#fulltext 9 | # 10 | # script to read gluster volume profile output retrieved every N seconds 11 | # and generate operation rate graph from it 12 | # 13 | # see gvp-README.html in the same directory for directions on use. 14 | # 15 | # Note: this tool uses a snapshot of javascript code from this project: 16 | # https://github.com/distributed-system-analysis/pbench 17 | # but we do not support any use of this software outside of the graphing 18 | # of the data generated below. The Save Image button does not work yet 19 | # 20 | # input: 21 | # this script expects input data to look like what the gvp.sh script (in same 22 | # directory) produces: 23 | # 24 | # record 1 is a timestamp in format YYYY-MM-DD-HH-MM 25 | # record 2 contains the user-specified sample interval and count 26 | # used by gvp.sh. 27 | # subsequent "gluster volume profile your-volume info" outputs are 28 | # concatenated to the profile log. 29 | # Each profile sample happens exactly N seconds after 30 | # the preceding sample, where N is the gvp.sh sampling interval. 31 | # seconds. The first sample happens N seconds after the timestamp. 32 | # 33 | # output: 34 | # 35 | # when we're all done reading in data, 36 | # we then print it out in a format suitable for spreadsheet-based graphing 37 | # 38 | # internals: 39 | # 40 | # the "intervals" array, indexed by interval number, stores results over time 41 | # within each array element, 42 | # we have a dictionary indexed by brick name containing BrickProfile instances 43 | # these in turn contain a dictionary of BrickFopProfile instances 44 | # to represent the per-FOP records in "gluster volume profile" output 45 | # 46 | # the per-brick dictionary is indexed by a string 47 | # starting with 'cumul' or 'intvl' and ending with the FOP name 48 | # this isn't strictly necessary but provides latent support 49 | # for someday including cumulative stats as well as per-interval stats 50 | # 51 | # stats for the entire volume are rolled up using call rates for weighted averaging 52 | # 53 | 54 | import sys 55 | import os 56 | from os.path import join 57 | import re 58 | import time 59 | import shutil 60 | import collections 61 | 62 | # fields in gluster volume profile output 63 | 64 | time_duration_types = ['cumulative', 'interval'] 65 | stat_names = ['pct-lat', 'avg-lat', 'min-lat', 'max-lat', 'call-rate'] 66 | directions = ['MBps-read', 'MBps-written'] 67 | min_lat_infinity = 1.0e24 68 | 69 | # this environment variable lets you graph .csv files using pbench 70 | 71 | pbench_graphs = True 72 | if os.getenv('SKIP_PBENCH_GRAPHING'): pbench_graphs = False 73 | 74 | # this is the list of graphs that will be produced 75 | 76 | graph_csvs = [ 77 | ('vol_call-rate_allfop', 'volume-level FOP call rates'), 78 | ('vol_pct-lat_allfop', 'percentage server-side latency by FOP'), 79 | ('MBps-written-volume', 'MB/sec written to Gluster volume'), 80 | ('MBps-read-volume', 'MB/sec read from Gluster volume'), 81 | ('MBps-written-bricks', 'MB/sec written to Gluster bricks'), 82 | ('MBps-read-bricks', 'MB/sec read from Gluster bricks') 83 | ] 84 | 85 | # all gvp.sh-generated profiles are expected to have these parameters 86 | # we define them here to have global scope, and they are only changed 87 | # by the input parser 88 | 89 | start_time = None 90 | expected_duration = None 91 | expected_sample_count = None 92 | sorted_fop_names = None 93 | sorted_brick_names = None 94 | intervals = None 95 | 96 | # this class stores per-fop statistics from gluster volume profile output 97 | # to compute stats for %latency and average latency across a set of bricks, 98 | # we have to compute averages weighted by brick usage. 99 | # We do this in two steps: 100 | # - loop over set of instances and compute weighted sum (not average) 101 | # - after loop, normalize using total calls 102 | 103 | 104 | class BrickFopProfile: 105 | 106 | def __init__(self, pct_lat, avg_lat, min_lat, max_lat, calls): 107 | self.pct_lat = pct_lat 108 | self.avg_lat = avg_lat 109 | self.min_lat = min_lat 110 | self.max_lat = max_lat 111 | self.calls = calls 112 | 113 | def __str__(self): 114 | return '%6.2f, %8.0f, %8.0f, %8.0f, %d' % ( 115 | self.pct_lat, self.avg_lat, self.min_lat, self.max_lat, self.calls) 116 | 117 | # append a single field to .csv record based on statistic type 118 | # use "-6.2f" instead of "%6.2f" so there are no leading spaces in record, 119 | # otherwise spreadsheet inserts colums at col. B 120 | 121 | def field2str(self, stat, duration): 122 | if stat == stat_names[0]: 123 | return '%-6.2f' % self.pct_lat 124 | elif stat == stat_names[1]: 125 | return '%8.0f' % self.avg_lat 126 | elif stat == stat_names[2]: 127 | if self.min_lat == min_lat_infinity: 128 | return '' # don't confuse spreadsheet/user 129 | else: 130 | return '%8.0f' % self.min_lat 131 | elif stat == stat_names[3]: 132 | if self.max_lat == 0: 133 | return '' 134 | else: 135 | return '%8.0f' % self.max_lat 136 | elif stat == stat_names[4]: 137 | call_rate = self.calls / float(duration) 138 | return '%9.2d' % call_rate 139 | 140 | # accumulate weighted sum of component profiles, will normalize them later 141 | 142 | def accumulate(self, addend): 143 | self.pct_lat += (addend.pct_lat * addend.calls) 144 | self.avg_lat += (addend.avg_lat * addend.calls) 145 | if addend.calls > 0: 146 | self.max_lat = max(self.max_lat, addend.max_lat) 147 | self.min_lat = min(self.min_lat, addend.min_lat) 148 | self.calls += addend.calls 149 | 150 | # normalize weighted sum to get averages 151 | 152 | def normalize_sum(self): 153 | try: 154 | # totals will become averages 155 | self.pct_lat /= self.calls 156 | self.avg_lat /= self.calls 157 | except ZeroDivisionError: # if no samples, set these stats to zero 158 | self.pct_lat = 0.0 159 | self.avg_lat = 0.0 160 | 161 | 162 | def zero_bfprofile(): 163 | # variable to accumulate stats across all bricks 164 | # for min, use some very large number 165 | # that latency will never exceed so that 166 | # min(lat, all_min_lat) == lat 167 | # same for max, use a lower bound for latency (0) 168 | # so max(lat, all_max_lat) = lat 169 | return BrickFopProfile(0.0, 0.0, min_lat_infinity, 0.0, 0) 170 | 171 | 172 | # this class stores per-brick results 173 | 174 | class BrickProfile: 175 | 176 | def __init__(self): 177 | self.bytes_read = 0 178 | self.bytes_written = 0 179 | self.interval = 0 # seconds, so DivisionByZero exception if not set 180 | # BrickFopProfile results stored in dictionary indexed by FOP name 181 | self.per_fop = {} 182 | 183 | def __str__(self): 184 | return '%d, %d, %s' % ( 185 | self.bytes_read, self.bytes_written, str(self.per_fop)) 186 | 187 | 188 | # if there is an error parsing the input... 189 | 190 | def usage(msg): 191 | print('ERROR: %s' % msg) 192 | print('usage: extract-glvolprof.py your-gluster-volume-profile.log') 193 | sys.exit(1) 194 | 195 | 196 | # because we produce so many .csv files, segregate them into a separate output 197 | # directory with pathname derived from the input log file with _csvdir suffix 198 | 199 | def make_out_dir(path): 200 | dir_path = path + '_csvdir' 201 | try: 202 | if os.path.exists(dir_path): 203 | shutil.rmtree(dir_path) 204 | os.mkdir(dir_path) 205 | except IOError: 206 | usage('could not (re-)create directory ' + dir_path) 207 | return dir_path 208 | 209 | 210 | # convert gluster volume profile output 211 | # into a time series of per-brick per-fop results. 212 | 213 | def parse_input(input_pathname): 214 | global start_time 215 | global expected_sample_interval 216 | global expected_sample_count 217 | global sorted_fop_names 218 | global sorted_brick_names 219 | global intervals 220 | 221 | try: 222 | with open(input_pathname, 'r') as file_handle: 223 | lines = [ l.strip() for l in file_handle.readlines() ] 224 | except IOError: 225 | usage('could not read ' + input_pathname) 226 | start_time = time.mktime(time.strptime(lines[0], '%Y-%m-%d-%H-%M')) * 1000 227 | tokens = lines[1].split() 228 | expected_sample_interval = int(tokens[0]) 229 | expected_sample_count = int(tokens[1]) 230 | print('collection started at %s' % lines[0]) 231 | print('sampling interval is %d seconds' % expected_sample_interval) 232 | print('expected sample count is %d samples' % expected_sample_count) 233 | 234 | # parse the file and record each cell of output in a way that lets you 235 | # aggregate across bricks later 236 | 237 | found_cumulative_output = False 238 | found_interval_output = False 239 | all_caps_name = re.compile('.*[A-Z]+$') 240 | fop_names = set() 241 | last_intvl = -1 242 | intvl = -1 243 | per_op_table = {} 244 | sample = -1 245 | intervals = [] 246 | bricks_seen = {} 247 | duration = None 248 | for ln in lines[2:]: 249 | tokens = ln.strip().split() 250 | 251 | if ln.startswith('Brick:'): 252 | 253 | brick_name = tokens[1] 254 | try: 255 | brick_count = bricks_seen[brick_name] 256 | except KeyError: 257 | brick_count = 0 258 | if brick_count == intvl + 1: 259 | intvl += 1 260 | else: 261 | assert brick_count == intvl 262 | brick_count += 1 263 | bricks_seen[brick_name] = brick_count 264 | 265 | elif ln.__contains__('Interval') and ln.__contains__('Stats'): 266 | 267 | assert intvl == last_intvl or intvl == last_intvl + 1 268 | last_intvl = intvl 269 | found_interval_output = True 270 | 271 | elif ln.__contains__('Cumulative Stats'): 272 | 273 | found_cumulative_output = True 274 | 275 | elif ln.__contains__('Duration:'): 276 | 277 | # we are at end of output for this brick and interval 278 | 279 | assert found_cumulative_output ^ found_interval_output 280 | duration = int(tokens[1]) 281 | if found_interval_output and \ 282 | abs(duration - expected_sample_interval) > 1: 283 | print(('WARNING: in sample %d brick %d the sample ' + 284 | 'interval %d deviates from expected value %d') % 285 | (brick_count, sample, duration, expected_sample_interval)) 286 | bricks_in_interval = intervals[intvl] 287 | brick = bricks_in_interval[brick_name] 288 | brick.interval = duration 289 | for fop in fop_names: 290 | for typ in time_duration_types: 291 | k = fop + '.' + typ 292 | try: 293 | bfprofile = brick.per_fop[k] 294 | except KeyError: 295 | bfprofile = zero_bfprofile() 296 | brick.per_fop[k] = bfprofile 297 | 298 | elif ln.__contains__('Data Read:'): 299 | 300 | bytes_read = int(tokens[2]) 301 | per_brick_info = bricks_in_interval[brick_name] 302 | per_brick_info.bytes_read = bytes_read 303 | 304 | elif ln.__contains__('Data Written'): 305 | 306 | bytes_written = int(tokens[2]) 307 | per_brick_info = bricks_in_interval[brick_name] 308 | per_brick_info.bytes_written = bytes_written 309 | 310 | # this is the end of per-brick results 311 | 312 | found_interval_output = False 313 | found_cumulative_output = False 314 | 315 | elif (found_interval_output or found_cumulative_output) \ 316 | and all_caps_name.match(ln): 317 | 318 | # we found a record we're interested in, 319 | # accumulate table of data for each gluster function 320 | 321 | sample += 1 322 | new_bfprofile = BrickFopProfile( 323 | float(tokens[0]), float(tokens[1]), float(tokens[3]), 324 | float(tokens[5]), int(tokens[7])) 325 | op_name = tokens[8] 326 | 327 | # op name is a key into dictionary, 328 | # we record both per-interval and cumulative stats 329 | 330 | fop_names.add(op_name) 331 | 332 | if found_interval_output: # keep cum. and interval stats separated 333 | op_name += '.' + time_duration_types[1] 334 | elif found_cumulative_output: 335 | op_name += '.' + time_duration_types[0] 336 | else: 337 | raise Exception('FOP-like string %s found outside stats' 338 | % op_name) 339 | 340 | if len(intervals) == intvl: 341 | bricks_in_interval = {} 342 | intervals.append(bricks_in_interval) 343 | elif len(intervals) == intvl + 1: 344 | bricks_in_interval = intervals[intvl] 345 | else: 346 | raise Exception(('intervals table length %d ' + 347 | 'does not match interval number %d') 348 | % (len(intervals), intvl)) 349 | 350 | try: 351 | fop_stats = bricks_in_interval[brick_name].per_fop 352 | except KeyError: 353 | bricks_in_interval[brick_name] = BrickProfile() 354 | fop_stats = bricks_in_interval[brick_name].per_fop 355 | 356 | fop_stats[op_name] = new_bfprofile 357 | 358 | sorted_brick_names = sorted(bricks_seen.keys()) 359 | sorted_fop_names = sorted(fop_names) 360 | return (start_time, intervals) 361 | 362 | 363 | # generate timestamp_ms column for pbench 364 | # given starting time of collection, sampling interval and sample number 365 | 366 | def gen_timestamp_ms(sample_index): 367 | return start_time + ((expected_sample_interval * sample_index) * 1000) 368 | 369 | 370 | # generate denominator for call rate computation based on duration type 371 | # can't use brick.interval 372 | 373 | def get_interval(duration_type, interval_index): 374 | if duration_type == 'cumulative': 375 | return interval_index * expected_sample_interval 376 | else: 377 | return expected_sample_interval 378 | 379 | # display bytes read and bytes written per brick and for entire volume 380 | # in separate graphs. If we put them in the same graph in a volume with 381 | # 16 bricks, for example, all you'll see is the per-volume number 382 | # normalize to MB/s with 3 decimal places so 1 KB/s/brick will show 383 | 384 | def gen_output_bytes(out_dir_path, duration_type): 385 | bytes_per_MB = 1000000.0 386 | final_brick_ct = len(sorted_brick_names) 387 | for direction in directions: 388 | per_vol_filename = direction + '-volume.csv' 389 | per_vol_pathname = join(out_dir_path, per_vol_filename) 390 | with open(per_vol_pathname, 'w') as total_transfer_fh: 391 | # when we support cumulative data, then we can name files this way 392 | #direction_filename = duration_type + '_' + direction + '.csv' 393 | per_brick_filename = direction + '-bricks.csv' 394 | per_brick_pathname = join(out_dir_path, per_brick_filename) 395 | with open(per_brick_pathname, 'w') as transfer_fh: 396 | if pbench_graphs: 397 | transfer_fh.write('timestamp_ms, ') 398 | total_transfer_fh.write('timestamp_ms, ') 399 | transfer_fh.write(','.join(sorted_brick_names)) 400 | total_transfer_fh.write('all\n') 401 | transfer_fh.write('\n') 402 | intvl = 0 403 | for bricks_in_interval in intervals: 404 | if pbench_graphs: 405 | transfer_fh.write('%d, ' % gen_timestamp_ms(intvl)) 406 | intvl += 1 407 | rate_interval = get_interval(duration_type, intvl) 408 | total_transfer = 0 409 | columns = [] 410 | for b in sorted_brick_names: # for each brick 411 | brick = bricks_in_interval[b] 412 | if direction.__contains__('read'): 413 | transfer = brick.bytes_read 414 | else: 415 | transfer = brick.bytes_written 416 | total_transfer += transfer 417 | columns.append( '%-8.3f ' % ((transfer/rate_interval)/bytes_per_MB)) 418 | transfer_fh.write(','.join(columns) + '\n') 419 | total_transfer_fh.write('%d, %-9.3f\n' % ( 420 | gen_timestamp_ms(intvl), 421 | (total_transfer/rate_interval)/bytes_per_MB)) 422 | 423 | 424 | # display per-FOP (file operation) stats, 425 | # both per brick and across all bricks 426 | 427 | def gen_per_fop_stats(out_dir_path, duration_type, stat): 428 | vol_fop_intervals = [] 429 | for fop in sorted_fop_names: 430 | #per_fop_filename = duration_type + '_' + stat + '_' + fop + '.csv' 431 | per_fop_filename = 'brick_' + stat + '_' + fop + '.csv' 432 | per_fop_path = join(out_dir_path, per_fop_filename) 433 | with open(per_fop_path, 'a') as fop_fh: 434 | hdr = '' 435 | if pbench_graphs: 436 | hdr += 'timestamp_ms, ' 437 | hdr += ','.join(sorted_brick_names) 438 | hdr += 'all\n' 439 | fop_fh.write(hdr) 440 | for i in range(0, len(intervals)): 441 | if pbench_graphs: 442 | fop_fh.write('%d, ' % gen_timestamp_ms(i)) 443 | bricks_in_interval = intervals[i] 444 | all_bfprofile = zero_bfprofile() 445 | columns = [] 446 | for b in sorted_brick_names: # for each brick 447 | brick = bricks_in_interval[b] 448 | try: 449 | fop_stats = brick.per_fop[fop + '.' + duration_type] 450 | except KeyError: 451 | fop_stats = zero_bfprofile() 452 | columns.append(fop_stats.field2str(stat, brick.interval)) 453 | all_bfprofile.accumulate(fop_stats) 454 | fop_fh.write('%s\n' % ','.join(columns)) 455 | 456 | # collect FOP results across all bricks for later 457 | 458 | all_bfprofile.normalize_sum() 459 | if len(vol_fop_intervals) == i: 460 | vol_fop_interval = {} 461 | vol_fop_intervals.append(vol_fop_interval) 462 | else: 463 | vol_fop_interval = vol_fop_intervals[i] 464 | vol_fop_interval[fop] = all_bfprofile 465 | return vol_fop_intervals 466 | 467 | def gen_fop_summary(dir_path, duration_type, stat, vol_fop_intervals): 468 | #vol_fop_profile_path = join(dir_path, duration_type + '_' + stat + '_allfop.csv') 469 | vol_fop_profile_path = join(dir_path, 'vol_' + stat + '_allfop.csv') 470 | with open(vol_fop_profile_path, 'w') as vol_fop_fh: 471 | if pbench_graphs: 472 | vol_fop_fh.write('timestamp_ms, ') 473 | vol_fop_fh.write(','.join(sorted_fop_names)) 474 | vol_fop_fh.write('\n') 475 | for i in range(0, len(vol_fop_intervals)): 476 | if pbench_graphs: 477 | vol_fop_fh.write('%d, ' % gen_timestamp_ms(i)) 478 | vol_fop_profile_interval = vol_fop_intervals[i] 479 | if duration_type == 'cumulative': 480 | sample_interval = (i + 1) * expected_sample_interval 481 | else: 482 | sample_interval = expected_sample_interval 483 | columns = [] 484 | for fop in sorted_fop_names: 485 | per_vol_fop_profile = vol_fop_profile_interval[fop] 486 | columns.append( 487 | per_vol_fop_profile.field2str( 488 | stat, sample_interval)) 489 | vol_fop_fh.write('%s\n' % ','.join(columns)) 490 | 491 | 492 | # generate graphs in 493 | # generate output files in separate directory from 494 | # data structure returned by parse_input 495 | 496 | next_graph_template=''' 497 |
    498 |

    %s 499 | 500 |
    501 |

    502 | 503 | 504 | 507 |
    508 | ''' 509 | 510 | def output_next_graph(graph_fh, gr_index): 511 | (csv_filename, graph_description) = graph_csvs[gr_index] 512 | gr_index += 1 # graph numbers start at 1 513 | graph_fh.write( next_graph_template % ( 514 | graph_description, gr_index, gr_index, csv_filename)) 515 | 516 | # static content of HTML file 517 | 518 | header=''' 519 | 520 | 521 | 522 | 523 | 524 | 525 | 526 | 527 | 528 | 529 | 530 | 531 | 532 | 533 |

    gluster volume profile - summary graphs

    534 | ''' 535 | 536 | trailer=''' 537 | 538 | 539 | ''' 540 | 541 | 542 | # generate graphs using header, trailer and graph template 543 | 544 | def gen_graphs(out_dir_path): 545 | graph_path = join(out_dir_path, 'gvp-graphs.html') 546 | with open(graph_path, 'w') as graph_fh: 547 | graph_fh.write(header) 548 | for j in range(0, len(graph_csvs)): 549 | output_next_graph(graph_fh, j) 550 | graph_fh.write(trailer) 551 | 552 | 553 | # make link to where javascript etc lives in unpacked tarball 554 | # ASSUMPTION is that output directory is a subdirectory of where this script 555 | # lives (not a sub-subdirectory). Sorry but that's the only way to generate a 556 | # softlink that works when we copy the csvdir to a different location. 557 | 558 | def gen_static_softlink(out_dir_path): 559 | saved_cwd = os.getcwd() 560 | static_dir = join(saved_cwd, 'static') 561 | if not os.path.exists(static_dir): 562 | print('ERROR: sorry, the javascript directory "static" ' + 563 | 'needs to be in same directory as this script, trying anyway...') 564 | os.chdir(out_dir_path) 565 | os.symlink(join('..', 'static'), 'static') 566 | os.chdir(saved_cwd) 567 | 568 | # generate everything needed to view the graphs 569 | 570 | def generate_output(out_dir_path): 571 | 572 | for t in [ 'interval' ]: # cumulative doesn't work yet 573 | gen_output_bytes(out_dir_path, t) 574 | for s in stat_names: 575 | vol_fop_intvls = gen_per_fop_stats(out_dir_path, t, s) 576 | gen_fop_summary(out_dir_path, t, s, vol_fop_intvls) 577 | 578 | gen_graphs(out_dir_path) 579 | gen_static_softlink(out_dir_path) 580 | 581 | sys.stdout.write('Gluster FOP types seen: ') 582 | for fop_name in sorted_fop_names: 583 | sys.stdout.write(' ' + fop_name) 584 | sys.stdout.write('\n') 585 | sys.stdout.write('Gluster bricks seen: ') 586 | for brick_name in sorted_brick_names: 587 | sys.stdout.write(' ' + brick_name) 588 | sys.stdout.write('\n') 589 | print('created Gluster statistics files in directory %s' % out_dir_path) 590 | print('graphs now available at browser URL file://%s/%s/gvp-graphs.html' \ 591 | % (os.getcwd(), out_dir_path)) 592 | 593 | 594 | # the main program is kept in a subroutine so that it can run on Windows. 595 | 596 | def main(): 597 | if len(sys.argv) < 2: 598 | usage('missing gluster volume profile output log filename parameter' 599 | ) 600 | fn = sys.argv[1] 601 | parse_input(fn) 602 | outdir = make_out_dir(fn) 603 | generate_output(outdir) 604 | 605 | main() 606 | --------------------------------------------------------------------------------