boost_server.php consumes 99% cpu

deu439356 · Post by **deu439356** » Wed Oct 08, 2008 4:29 pm

I am not seeing any other errors and I am using the lastest version of Boost from the SVN. Do you know if there is a new poller that is supposed to be installed with boost? I am running 8.7b

khufure · Post by **khufure** » Wed Oct 08, 2008 4:29 pm

khufure wrote:What version of boost? You should be using the Subversion one. Latest stable (1.7) is not really usable.

Are there any messages about boost around the 2006 errors? E.g.
Code: Select all
ERROR: Timeout detected.  Boost server is down.  Contact support immediately!!

Nevermind, I see in your log similar messages do exist. Checkout the last 2 pages of http://forums.cacti.net/viewtopic.php?t=18430 for a way to restart the boost server if/when it dies.

Post by **TheWitness** » Wed Oct 08, 2008 7:20 pm

I would make it an inittab process.

TheWitness

deu439356 · Post by **deu439356** » Thu Oct 09, 2008 5:07 pm

Ok. We turned the multiprocess server back on to help write the RRDs faster when the boost poller runs every hour (which we have to force). It seems that the mysql errors (the 2006 error) was being caused by boost getting backed up. We have also increased the allowed run time of boost so it has enough time to write to the RRDs. The last time boost ran, it took close to 1400 seconds. My next question is: why didn't the poller time decrease when we installed boost?

SYSTEM STATS: Time:265.0897 Method:spine Processes:15 Threads:15 Hosts:1025 HostsPerProcess:69 DataSources:120435 RRDsProcessed:0

Thanks,

Post by **TheWitness** » Thu Oct 09, 2008 6:44 pm

Please provide some details from the Tech Support page. Only the counts of what is being polled, I don't care about the various PHP modules and their settings, besides, you shouldn't post that here anyway.

The reason that the polling time went down in general is that you don't have to wait on Disk I/O.

The solution to your problem is multi-fold. We can discuss that more when I have additional information from you.

TheWitness

deu439356 · Post by **deu439356** » Fri Oct 10, 2008 9:43 am

Code: Select all

Technical Support
 
[b]General Information [/b]
Date Fri, 10 Oct 2008 09:27:10 -0500 
Cacti Version 0.8.7b 
Cacti OS unix 
SNMP Version net-snmp 
RRDTool Version RRDTool 1.2.x 
Hosts 1387 
Graphs 40447 
Data Sources:
                  Script/Command: 2
                  SNMP: 2932
                  SNMP Query: 39000
                  Script - Script Server (PHP): 4
                  Script Query - Script Server: 124
                  Total: 42062 



[b]Poller Information [/b]
Interval 300 
Type spine 
Items: 
                  Action[0]: 120219
                  Action[1]: 2
                  Action[2]: 252
                  Total: 120473 

Concurrent Processes 15 
Max Threads 15 
PHP Servers 10 
Script Timeout 45 
Max OID 20 
Last Run Statistics Time:251.7290 Method:spine Processes:15 Threads:15 Hosts:1035 HostsPerProcess:69 DataSources:120471 RRDsProcessed:0

dononeil · Post by **dononeil** » Wed Oct 29, 2008 1:24 pm

I'm experiencing a similar problem... boost server chews up all the CPU it can get after running for 24 hours. Restarting it seems to solve the issue.

I'm running the latest SVN as of 10/18/08.

Post by **TheWitness** » Wed Oct 29, 2008 7:08 pm

Could be rrdtool segmentation faults causing the race condition. Try going using cacti 0.8.7c-beta1 and see if the problem is resolved.

TheWitness

dononeil · Post by **dononeil** » Mon Nov 03, 2008 12:06 pm

I can't just 'try going to beta' code to see if this fixes the problem without some more specific indications as to where the problem is coming from.

I could however upgrade to the latest RRDtool if you think that would make a difference.

khufure · Post by **khufure** » Fri Nov 07, 2008 5:05 pm

dononeil wrote:I can't just 'try going to beta' code to see if this fixes the problem without some more specific indications as to where the problem is coming from.

I could however upgrade to the latest RRDtool if you think that would make a difference.

rrdtool-1.3.x will segfault. you could try 1.2.27 or 1.2.28G (or whatever they are calling the version with the font patch).

dononeil · Post by **dononeil** » Fri Nov 07, 2008 5:07 pm

You sure they haven't fixed it in the latest 1.3.4 version just released last month?

ceyounger · Post by **ceyounger** » Sat Nov 08, 2008 5:19 am

Throwing my hat into the ring.

I have the same errors listed here. Had to disable on-demand updating, which defeats the purpose of this plugin.

I implemented all the steps listed in this thread and everytime I turn on on-demand updating I get the 2006 errors.

dononeil · Post by **dononeil** » Mon Nov 10, 2008 2:37 pm

Ok... I built rrdtool 1.2.28 and as soon as I replace the original 1.2.15 binaries I have on my system with those the text rendering breaks (no text at all).

Any ideas what could be causing this?

khufure · Post by **khufure** » Wed Nov 12, 2008 6:02 pm

dononeil wrote:Ok... I built rrdtool 1.2.28 and as soon as I replace the original 1.2.15 binaries I have on my system with those the text rendering breaks (no text at all).

Any ideas what could be causing this?

A bug in rrdtool-1.2.28. Get the patch or use 1.2.27.

dononeil · Post by **dononeil** » Wed Nov 19, 2008 5:07 pm

Ok... I've patched rrdtool 1.28 as well as tried rrdtool 1.27 and I'm still seeing the same issue. See the attached image. It seems to happen after 24 hours as you can see.

boost_server.php consumes 99% cpu

Who is online