RRDTool 1.2.x utilizes 100% cpu

[BSOD2600]

I've given up compiling my own rrdtool and instead used the one hosted on this site for now.

The new rrdtool 1.2.8 apparently has no problems updating my rra files (I see the output of cmd.php with numbers). But the problem I have is when trying to view any graph. Soon as I click on the graph tab in cacti, several rrdtool.exe processes open up. Each consumes a portion of the cpu, in all totaling 100%. This has been the case for the last several rrdtool versions. I never even get the blank image icons. I can kill off these processes easily in task manager. The new rrdtool has the same permissions as the older 1.0.x which works.

If I go back to 1.2.1, which TheWitness sent me, something a little different happens. When I try to view the graphs, I immedatly get the blank image icons. When in graph management, attempting to view a graph, I click on turn on graph debug mode. I get a the broken image icon. This rrdtool.exe does not suck down all the cpu. This rrdtool.exe is running under the 'network service' user account (instead of mine, which the normal cacti polling process runs under). I am unable to kill it either in task manager (says access denied). Only way I can kill it, is by restarting IIS.

If I go back to 1.0.49, everything works great.

What the heck is broken now?

[TheWitness]

Which graphs are locking up? You need to send me the graph debug informaiton. Set logging to DEBUG, then view the specific graph that breaks from the Graphs option in cacti. Then goto the cacti log and cut the cacti2rrd graph command. Send me both it and the RRD file in question so I can send to Tobi for investigation.

TheWitness

[BSOD2600]

All graphs lock up.

I do recall, when rrdtool 1.2.1 came out, I was able to see the graphs just fine (remember me commenting about the larger png file sizes). Unless my roommate has inadvertently changed some security setting somewhere else on the system, everything is the same as before. Since rrdtool 1.2.x doesn't work now, something is obviously broken.

Oh yea, I finally did get rrdtool to compile properly on my system. I did a complete reinstall of Cygwin. Now it builds great.

[BSOD2600]

Alright, I got to looking at the cacti2rrd outputs.

With rrdtool 1.0 selected in cacti and using rrdtool 1.0.49, this is the output and it works:

Code: Select all

WEBLOG: Poller[0] CACTI2RRD: c:/rrdtool/rrdtool.exe graph -  --imgformat=PNG  --start=1116443250  --end=1116529650  --title="WebSTAR DPX2100 Cable Modem - DOCSIS Diagnostics #2"  --base=1000  --height=120  --width=500  --alt-autoscale-max  --lower-limit=0  COMMENT:"From 2005/05/18 12:07:30 To 2005/05/19 12:07:30\c"  COMMENT:"  \n"  --vertical-label=""  DEF:a="c\:\auntie\cacti/rra/webstar_dpx2100_cable_modem_docsis_lostsyncs_2.rrd":docsis_lostsyncs:AVERAGE  DEF:b="c\:\auntie\cacti/rra/webstar_dpx2100_cable_modem_docsis_resets_2.rrd":docsis_resets:AVERAGE  DEF:c="c\:\auntie\cacti/rra/webstar_dpx2100_cable_modem_docsis_t3_2.rrd":docsis_t3:AVERAGE  DEF:d="c\:\auntie\cacti/rra/webstar_dpx2100_cable_modem_docsis_t4_2.rrd":docsis_t4:AVERAGE  LINE2:a#0000FF:"Lost Syncs"   GPRINT:a:LAST:" Current\:%8.2lf %s"   GPRINT:a:AVERAGE:"Avg\:%8.2lf %s"   GPRINT:a:MIN:"Min\:%8.2lf %s"   GPRINT:a:MAX:"Max\:%8.2lf %s\n"   LINE2:b#FF5700:"Resets"   GPRINT:b:LAST:"     Current\:%8.2lf %s"   GPRINT:b:AVERAGE:"Avg\:%8.2lf %s"   GPRINT:b:MIN:"Min\:%8.2lf %s"   GPRINT:b:MAX:"Max\:%8.2lf %s\n"   LINE2:c#837C04:"T3 Timeouts"   GPRINT:c:LAST:"Current\:%8.2lf %s"   GPRINT:c:AVERAGE:"Avg\:%8.2lf %s"   GPRINT:c:MIN:"Min\:%8.2lf %s"   GPRINT:c:MAX:"Max\:%8.2lf %s\n"   LINE2:d#6EA100:"T4 Timeouts"   GPRINT:d:LAST:"Current\:%8.2lf %s"   GPRINT:d:AVERAGE:"Avg\:%8.2lf %s"   GPRINT:d:MIN:"Min\:%8.2lf %s"   GPRINT:d:MAX:"Max\:%8.2lf %s\n" 

When I tried it with rrdtool 1.2 selected in cacti and using 1.2.8, this is the output which did not work:

Code: Select all

WEBLOG: Poller[0] CACTI2RRD: c:/rrdtool/rrdtool1.2.8.exe graph -  --imgformat=PNG  --start=1116443668  --end=1116530068  --title="WebSTAR DPX2100 Cable Modem - DOCSIS Diagnostics #2"  --base=1000  --height=120  --width=500  --alt-autoscale-max  --lower-limit=0  COMMENT:"From 2005/05/18 12\:14\:28 To 2005/05/19 12\:14\:28\c"  COMMENT:"  \n"  --vertical-label=""  --slope-mode  DEF:a="c\:\auntie\cacti/rra/webstar_dpx2100_cable_modem_docsis_lostsyncs_2.rrd":docsis_lostsyncs:AVERAGE  DEF:b="c\:\auntie\cacti/rra/webstar_dpx2100_cable_modem_docsis_resets_2.rrd":docsis_resets:AVERAGE  DEF:c="c\:\auntie\cacti/rra/webstar_dpx2100_cable_modem_docsis_t3_2.rrd":docsis_t3:AVERAGE  DEF:d="c\:\auntie\cacti/rra/webstar_dpx2100_cable_modem_docsis_t4_2.rrd":docsis_t4:AVERAGE  LINE2:a#0000FF:"Lost Syncs"   GPRINT:a:LAST:" Current\:%8.2lf %s"   GPRINT:a:AVERAGE:"Avg\:%8.2lf %s"   GPRINT:a:MIN:"Min\:%8.2lf %s"   GPRINT:a:MAX:"Max\:%8.2lf %s\n"   LINE2:b#FF5700:"Resets"   GPRINT:b:LAST:"     Current\:%8.2lf %s"   GPRINT:b:AVERAGE:"Avg\:%8.2lf %s"   GPRINT:b:MIN:"Min\:%8.2lf %s"   GPRINT:b:MAX:"Max\:%8.2lf %s\n"   LINE2:c#837C04:"T3 Timeouts"   GPRINT:c:LAST:"Current\:%8.2lf %s"   GPRINT:c:AVERAGE:"Avg\:%8.2lf %s"   GPRINT:c:MIN:"Min\:%8.2lf %s"   GPRINT:c:MAX:"Max\:%8.2lf %s\n"   LINE2:d#6EA100:"T4 Timeouts"   GPRINT:d:LAST:"Current\:%8.2lf %s"   GPRINT:d:AVERAGE:"Avg\:%8.2lf %s"   GPRINT:d:MIN:"Min\:%8.2lf %s"   GPRINT:d:MAX:"Max\:%8.2lf %s\n" 

When I tried that command manually, this was the output which never ended: couldn't transform glyph.

Seems this is why the rrdtool processes would suck down the cpu and never end!

When I tried the rrdtool 1.0 output with rrdtool 1.2.1, it would die right away with: ERROR: unknown option '--slope-mode'. When I tried the rrdtool 1.2 output with rrdtool 1.2.1 it would die with: ERROR: Garbage ':07:30 To 2005/05/19 12:07:30\c' after command: COMMENT:From 2005/05/18 12:07:30 To 2005/05/19 12:07:30\c

[TheWitness]

Please goto Settings->General and change your rrdtool version to 1.0.x

TheWitness

[BSOD2600]

huh? I did. When I said, "With rrdtool 1.0 selected in cacti" or "When I tried it with rrdtool 1.2 selected in cacti", that is what I was referring to. Going to Settings->General and changing the rrdtool version.

[BSOD2600]

Fixed!

I changed the RRD_DEFAULT_FONT variable from lucon.ttf (like the document originally suggested.... yes, the file does exist) to arial.ttf. Problem was fixed. I also tested this on my WinXP Pro SP2 box. Using lucon also would cause the glyph error. Changing it to a different font, or simply not using RRD_DEFAULT_FONT would fix the problem.

Still sounds like a bug in rrdtool to go into an endless loop with that error output.

[bulek]

I also can confirm that using Lucida font on Win2000 server goes to endless loop.

- Piotr