<html> <head> <link rel=stylesheet href="style.css" type="text/css"> <title>Collectl Examples</title> </head> <body> <center><h1>Examples</h1></center> <p> There are far too many combinations of switches and output formats so only a few of the more basic ones will be shown below. These examples show both the command and in most cases the resultant output. For more examples see both the <a href=http://collectl.sourceforge.net/FAQ-collectl.html>FAQ</a> and collectl man page after you install it. <p> <h2>Interactive Commands</h2> The following examples show the results of running collectl interactively and seeing system performance numbers in real-time. <p> <h3>Default</h3> Notice that in this mode you see one line per sampling interval. You are only limited by the width of your terminal window. <div class=terminal> <pre> [root@poker]# collectl #<-------CPU--------><-----------Disks-----------><-----------Network----------> #cpu sys inter ctxsw KBRead Reads KBWrit Writes netKBi pkt-in netKBo pkt-out 0 0 134 30 0 0 0 0 0 1 0 1 0 0 136 39 0 0 200 3 2 20 0 4 0 0 130 30 0 0 0 0 0 1 0 0 2 2 134 24 0 0 0 0 2 18 0 2 </pre> </div> <h3>Default, but in verbose mode</h3> In this mode you give up one line per interval and are rewarded with more details than could fit on a single line. This format always includes the date and time. <div class=terminal> <pre> [root@poker]# collectl --verbose ### RECORD 1 >>> cag-dl380-01 <<< (1179493640.005) (Fri May 18 09:07:20 2007) ### # CPU SUMMARY (INTR, CTXSW & PROC /sec) # USER NICE SYS WAIT IRQ SOFT STEAL IDLE INTR CTXSW PROC RUNQ RUN AVG1 AVG5 AVG15 1 0 0 0 0 0 0 98 354 501 1 100 1 0.41 0.12 0.04 # DISK SUMMARY (/sec) #KBRead RMerged Reads SizeKB KBWrite WMerged Writes SizeKB 0 0 0 0 0 0 0 0 # NETWORK SUMMARY (/sec) # KBIn PktIn SizeIn MultI CmpI ErrIn KBOut PktOut SizeO CmpO ErrOut 17 121 149 0 0 0 15 88 175 0 0 </pre> </div> <h3>Detail Data</h3> If you would rather see details on specific instances, use the uppercase subsystem names with -s, so rather than the default of <i>-scdn</i> use <i>-sCDN</i>, which also forces --verbose, noting that you can also mix lower and uppercase subsystem types. <div class=terminal-wide15> <pre> [root@poker]# collectl -sCDN ### RECORD 1 >>> cag-dl380-01 <<< (1179493735.005) (Fri May 18 09:08:55 2007) ### # SINGLE CPU STATISTICS # CPU USER NICE SYS WAIT IRQ SOFT STEAL IDLE 0 0 0 0 0 0 0 0 100 1 0 0 0 0 0 0 0 99 # DISK STATISTICS (/sec) # <---------reads---------><---------writes---------><--------averages--------> Pct #Name KBytes Merged IOs Size KBytes Merged IOs Size RWSize QLen Wait SvcTim Util cciss/c0d0 0 0 0 0 0 0 0 0 0 0 0 0 0 cciss/c0d1 0 0 0 0 0 0 0 0 0 0 0 0 0 cciss/c0d2 0 0 0 0 0 0 0 0 0 0 0 0 0 # NETWORK STATISTICS (/sec) #Num Name KBIn PktIn SizeIn MultI CmpI ErrIn KBOut PktOut SizeO CmpO ErrOut 0 lo: 0 0 0 0 0 0 0 0 0 0 0 1 eth0: 0 2 207 0 0 0 0 0 0 0 0 2 eth1: 0 2 207 0 0 0 0 0 0 0 0 3 eth2: 1 20 72 0 0 0 0 4 122 0 0 4 eth3: 0 0 0 0 0 0 0 0 0 0 0 </pre> </div> <h3>Mixed Summary and Detail Data</h3> For brevity we're only showing cpu and disk data. Note that we can show both cpu summary as well as detail while we're only showing disk details. <div class=terminal-wide14> <pre> [root@poker]# collectl -scCD ### RECORD 1 >>> cag-dl380-01 <<< (1192729823.010) (Thu Oct 18 13:50:23 2007) ### # CPU SUMMARY (INTR, CTXSW & PROC /sec) # USER NICE SYS WAIT IRQ SOFT STEAL IDLE INTR CTXSW PROC RUNQ RUN AVG1 AVG5 AVG15 0 0 0 0 0 0 0 99 135 30 0 145 0 0.01 0.01 0.00 # SINGLE CPU STATISTICS # CPU USER NICE SYS WAIT IRQ SOFT STEAL IDLE 0 0 0 0 0 0 0 0 100 1 0 0 0 0 0 0 0 99 # DISK STATISTICS (/sec) # <---------reads---------><---------writes---------><--------averages--------> Pct #Name KBytes Merged IOs Size KBytes Merged IOs Size RWSize QLen Wait SvcTim Util cciss/c0d0 0 0 0 0 0 0 0 0 0 0 0 0 0 cciss/c0d1 0 0 0 0 0 0 0 0 0 0 0 0 0 cciss/c0d2 0 0 0 0 0 0 0 0 0 0 0 0 0 </pre> </div> <h3>Different Subsystems With Timestamps</h3> <div class=terminal> <pre> [root@poker]# collectl -scft -oT waiting for 1 second sample... # <--------CPU--------><------------TCP-------------><------NFS Totals------> #Time cpu sys inter ctxsw PureAcks HPAcks Loss FTrans read write meta comm 08:12:20 0 0 1007 120 1 0 0 0 0 0 0 0 08:12:21 1 1 1077 400 1 0 0 0 0 0 0 0 </pre> </div> <h3>When you just don't know...</h3> A great way to familiarize yourself with the types of data collectl can generate is to do (as of Version 2.6.0) <i>--all --verbose</i> and see all the summary data generated at once, excluding processes and slabs. Since this can be a lot to watch, especially as it scrolls off the screen between samples, the <i>--home</i> can be your friend. It will clear the screen between samples and remove extra lines to give an appearance of a continously refreshing screen-based utility. Try it with different combinations of subsystems to reduce the amount of information displayed. <div class=terminal-wide14> <pre> ### RECORD 1 >>> hadesn1 <<< (1214918640.001) (Tue Jul 1 09:24:00 2008) ### # CPU SUMMARY (INTR, CTXSW & PROC /sec) # USER NICE SYS WAIT IRQ SOFT STEAL IDLE INTR CTXSW PROC RUNQ RUN AVG1 AVG5 AVG15 0 0 0 0 0 0 0 100 1327 647 1 341 0 0.00 0.00 0.00 # INTERRUPT SUMMARY # Cpu0 Cpu1 Cpu2 Cpu3 Cpu4 Cpu5 Cpu6 Cpu7 999 0 300 0 0 0 26 0 # DISK SUMMARY (/sec) #KBRead RMerged Reads SizeKB KBWrite WMerged Writes SizeKB 0 0 0 0 0 0 0 0 # NFS SUMMARY (/sec) #<---------------------------server---------------------------><----------------client----------------> # Reads Writes Meta Comm UDP TCP TCPConn BadAuth BadClnt Reads Writes Meta Comm Retrans Authref 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 # INODE SUMMARY # Dentries File Handles Inodes # Number Unused Alloc % Max Number 42532 39837 510 0.03 39011 # LUSTRE CLIENT SUMMARY # KBRead Reads KBWrite Writes 0 0 0 0 # MEMORY STATISTICS #<------------------------Physical Memory-----------------------><-----------Swap----------><-Inactive-><Pages/sec> # TOTAL USED FREE BUFF CACHED SLAB MAPPED COMMIT TOTAL USED FREE TOTAL IN OUT 16053M 1276M 14776M 130816K 403696K 587884K 80992K 130684K 15308M 0 15308M 201028K 0 0 # NETWORK SUMMARY (/sec) # KBIn PktIn SizeIn MultI CmpI ErrIn KBOut PktOut SizeO CmpO ErrOut 1 15 103 0 0 0 0 2 150 0 0 # SOCKET STATISTICS # <-------------Tcp-------------> Udp Raw <---Frag--> #Used Inuse Orphan Tw Alloc Mem Inuse Inuse Inuse Mem 90 8 0 1 11 0 8 0 0 0 # TCP SUMMARY (/sec) # PureAcks HPAcks Loss FTrans 0 1 0 0 # INFINIBAND SUMMARY (/sec) # KBIn PktIn SizeIn KBOut PktOut SizeOut Errors 0 0 0 0 0 0 0 </pre></div> <p> <h2>Record Mode</h2> This mode is often used when running a test of a limited duration from a couple of minutes to several hours or more and since collectl is not being run as a daemon, the default sampling rate is 1 second. <p> <h3>Collect 100 Samples and Exit</h3> <div class=terminal> <pre> [root@poker]# collectl -c100 -f/tmp </pre> </div> <h3>Run Until Terminated With ^C</h3> <div class=terminal> <pre> [root@poker]# collectl -f/tmp </pre> </div> <p> <h2>Playback Mode</h2> There are a couple of things to remember about playback: <ul> <li>You need a data file. When run as a deamon via <i>service collectl start</i> data will be written to /var/log/collectl by default.</li> <li>You can playback a subset of data by inclusion of the -s switch</li> <li>Just about any formatting switches you can use interactively you can use during playback</li> <li>You can select a subset of the time period with --from and --thru</li> </ul> <h3>Playback Between 2 Time Periods</h3> In this example we're not selecting any device details and so the output defaults to brief mode and all data is printing on the same line. We've also chosen to display time in msecs and use --from to specify both times. <div class=terminal-wide15> <pre> collectl -scdn -p /var/log/collectl/cag-dl380-01-20070830-082013.raw.gz --from 08:29-08:30 -oTm # <--------CPU--------><-----------Disks-----------><-----------Network----------> #Time cpu sys inter ctxsw KBRead Reads KBWrit Writes netKBi pkt-in netKBo pkt-out 08:29:00.012 0 0 135 38 0 0 2 0 0 11 0 2 08:29:10.012 2 0 142 142 0 0 142 2 1 14 1 5 08:29:20.012 1 0 138 45 0 0 33 1 1 14 0 3 08:29:30.012 0 0 135 52 0 0 5 0 1 11 0 3 08:29:40.012 0 0 136 44 0 0 21 0 1 11 0 3 08:29:50.012 1 0 177 123 14 2 385 38 1 13 1 4 </pre> </div> <h3>Same Data File, Shorter Interval, Disk Details</h3> As expected this defaults to --verbose mode. Also notice we left off the leading 0 in the <i>from time</i>. <div class=terminal-wide14> <pre> collectl -sD -p /var/log/collectl/cag-dl380-01-20070830-082013.raw.gz -from 8:29 --thru 08:29:10 -oD # DISK STATISTICS (/sec) # <---------reads---------><---------writes---------><--------averages--------> Pct # Name KBytes Merged IOs Size KBytes Merged IOs Size RWSize QLen Wait SvcTim Util 08:29:00 c0d0 0 0 0 0 2 0 0 6 6 0 6 6 0 08:29:00 c0d1 0 0 0 0 0 0 0 0 0 0 0 0 0 08:29:00 c0d2 0 0 0 0 0 0 0 0 0 0 0 0 0 </pre> </div> <h3>Same Data File, Different Data</h3> Notice here that we're using -oD instead of -T and so both the date and time are displayed. We could have also chosen to force the time in msecs but chose not to in order to save screen real estate. <div class=terminal-wide15> <pre> collectl -sms -p /var/log/collectl/cag-dl380-01-20070830-082013.raw.gz --from 08:29-08:30 -oD # <-----------Memory----------><------Sockets-----> #Date Time free buff cach inac slab map Tcp Udp Raw Frag 20070830 08:29:00 64M 529M 2G 483M 0 0 89 17 0 0 20070830 08:29:10 64M 529M 2G 483M 0 0 89 17 0 0 20070830 08:29:20 64M 529M 2G 483M 0 0 89 17 0 0 20070830 08:29:30 64M 529M 2G 483M 0 0 89 17 0 0 20070830 08:29:40 64M 529M 2G 483M 0 0 89 17 0 0 20070830 08:29:50 64M 529M 2G 483M 0 0 89 17 0 0 </pre> </div> <h3>Same As Last Command But Display In Plot Format</h3> This format is rarely used, but nothing prevents one from doing so. <div class=terminal-wide12> <pre> [root@poker]# collectl -P -sms -p /var/log/collectl/cag-dl380-01-20070830-082013.raw.gz --from 08:29-08:30 #Date Time [MEM]Tot [MEM]Used [MEM]Free [MEM]Shared [MEM]Buf [MEM]Cached [MEM]Slab [MEM]Map [MEM]Commit [MEM]SwapTot [MEM]SwapUsed [MEM]SwapFree [MEM]Dirty [MEM]Clean [MEM]Laundry [MEM]Inactive [MEM]PageIn [MEM]PageOut [SOCK]Used [SOCK]Tcp [SOCK]Orph [SOCK]Tw [SOCK]Alloc [SOCK]Mem [SOCK]Udp [SOCK]Raw [SOCK]Frag [SOCK]FragMem 20070830 08:29:00 3098632 3032208 66424 0 542188 2258744 0 0 0 2044056 36596 2007460 336 62176 432280 494792 0 2 89 39 0 0 39 2 17 0 0 0 20070830 08:29:10 3098632 3032412 66220 0 542244 2258744 0 0 0 2044056 36596 2007460 1208 62176 431404 494788 0 142 89 39 0 0 39 1 17 0 0 0 20070830 08:29:20 3098632 3032440 66192 0 542276 2258776 0 0 0 2044056 36596 2007460 1412 62176 431232 494820 0 34 89 39 0 0 39 2 17 0 0 0 20070830 08:29:30 3098632 3032464 66168 0 542292 2258776 0 0 0 2044056 36596 2007460 1412 62176 431232 494820 0 5 89 39 0 0 39 2 17 0 0 0 20070830 08:29:40 3098632 3032464 66168 0 542300 2258776 0 0 0 2044056 36596 2007460 1584 62176 431060 494820 0 22 89 39 0 0 39 2 17 0 0 0 20070830 08:29:50 3098632 3033020 65612 0 542532 2258700 0 0 0 2044056 36596 2007460 2504 61628 430748 494880 14 386 89 39 0 0 39 1 17 0 0 0 </pre> </div> <h3>Playback and Convert To Non-Compressed Plot File</h3> There's really not much to see here since the only ouput this command will produce is error messages. <div class=terminal-wide15> <pre> [root@poker]# collectl -p /var/log/collectl/cag-dl380-01-20070830-082013.raw.gz -P -f /tmp -oz </pre> </div> </body> </html>