|
|
| Author |
Message |
tripleK
Joined: 02 Sep 2008
Posts: 4
|
 Posted: Tue Sep 02, 2008 2:22 pm Post subject: CPU usage issue & Error |
 |
|
Hi,
I am encountering an issue regarding the CPU usage of my cacti server.
The server CPU usage was push to 100% while the polling process is running and also when I am viewing the graph on the web UI the CPU usage of my server also achieve 100% CPU usage.
I would like to ask about is this issue normal? Is there a way to reduce the CPU usage?
This is the specification of my server:
Dual Processor Xeon 3.0GHz Processor.
2GB RAM
Windows Server 2003 Enterprise Edition with Service Pack 2
Cacti version 0.8.7b
PHP 5.2.3
Mysql 5.0.51a
IIS 6.0
I am running on cmd.php for polling process.
09/02/2008 07:10:21 PM - SYSTEM STATS: Time:21.3391 Method:cmd.php Processes:1 Threads:N/A Hosts:25 HostsPerProcess:25 DataSources:1604 RRDsProcessed:802
Besides, my polling process have this error message. How could I solve the problem?
09/02/2008 07:10:13 PM - CMDPHP: Poller[0] Host[39] DS[817] WARNING: Result from SNMP not valid. Partial Result:
09/02/2008 07:10:07 PM - CMDPHP: Poller[0] Host[14] DS[688] WARNING: Result from SNMP not valid. Partial Result:
09/02/2008 07:10:07 PM - CMDPHP: Poller[0] Host[14] DS[688] WARNING: Result from SNMP not valid. Partial Result:
09/02/2008 07:10:07 PM - CMDPHP: Poller[0] Host[14] DS[687] WARNING: Result from SNMP not valid. Partial Result:
09/02/2008 07:10:07 PM - CMDPHP: Poller[0] Host[14] DS[687] WARNING: Result from SNMP not valid. Partial Result:
09/02/2008 07:10:07 PM - CMDPHP: Poller[0] Host[14] DS[685] WARNING: Result from SNMP not valid. Partial Result:
09/02/2008 07:10:07 PM - CMDPHP: Poller[0] Host[14] DS[685] WARNING: Result from SNMP not valid. Partial Result:
09/02/2008 07:10:07 PM - CMDPHP: Poller[0] Host[14] DS[684] WARNING: Result from SNMP not valid. Partial Result:
09/02/2008 07:10:07 PM - CMDPHP: Poller[0] Host[14] DS[684] WARNING: Result from SNMP not valid. Partial Result:
Thanks!!
Best regards,
Kenny |
|
| Back to top |
|
 |
BSOD2600
Cacti Moderator
Joined: 08 May 2004
Posts: 6763
Location: WA, USA
|
| Posted: Tue Sep 02, 2008 5:37 pm Post subject: Re: CPU usage issue & Error |
|
|
| tripleK wrote: | | I would like to ask about is this issue normal? Is there a way to reduce the CPU usage? |
Yes, this is normal. Look in taskman, what processes are using the cpu? If its php, mysql and rrdtool, this is expected. I don't know of a way to lower the max amount of cpu used...
| tripleK wrote: | | Besides, my polling process have this error message. How could I solve the problem? |
Look in the snmp poller cache for those various DS numbers (or you can filter by the host #). Manually check the OIDs which cacti is using are correct (i.e. snmpwalk or getif them). For the majority of them, proably the devices just need to be reindexed so the OIDs are refreshed. you can also clear the poller cache, since some times that helps with this issue. |
|
| Back to top |
|
|
rzpr01
Joined: 14 May 2008
Posts: 2
|
| Posted: Tue Sep 02, 2008 9:25 pm Post subject: server CPU usage was push to 100% |
|
|
I encountered the same problem with more less spec. hw and w/ w2k3 ent. ed. sp1 a few days ago.
But now, i hope solved...the way i did was to change :
Tab Graph Export : Export Methode --> disabled and Export Timing --> disabled
(Console>configuration>setting) |
|
| Back to top |
|
|
tripleK
Joined: 02 Sep 2008
Posts: 4
|
| Posted: Wed Sep 03, 2008 10:28 am Post subject: Re: CPU usage issue & Error |
|
|
Hi, thanks for replying.
| BSOD2600 wrote: | Yes, this is normal. Look in taskman, what processes are using the cpu? If its php, mysql and rrdtool, this is expected. I don't know of a way to lower the max amount of cpu used...
|
As for the cpu usage, the processor is still ok which reached about 40% of the CPU usage while the polling process is running. But whenever I was viewing the graph, the CPU usage is 100% and the process that causing it to 100% is w3wp.exe. What is the actual process that was running in the background? Is there a way to reduce the processor usage while viewing the graph. I do monitor the graph very frequently and with multiple user viewing at it as well.
| BSOD2600 wrote: |
Look in the snmp poller cache for those various DS numbers (or you can filter by the host #). Manually check the OIDs which cacti is using are correct (i.e. snmpwalk or getif them). For the majority of them, proably the devices just need to be reindexed so the OIDs are refreshed. you can also clear the poller cache, since some times that helps with this issue. |
I did a snmpwalk on those DS and the snmpwalk return no results. How can I do reindex the OID? Will it cause my previouse data lost? |
|
| Back to top |
|
|
tripleK
Joined: 02 Sep 2008
Posts: 4
|
| Posted: Wed Sep 03, 2008 10:32 am Post subject: Re: server CPU usage was push to 100% |
|
|
| rzpr01 wrote: | I encountered the same problem with more less spec. hw and w/ w2k3 ent. ed. sp1 a few days ago.
But now, i hope solved...the way i did was to change :
Tab Graph Export : Export Methode --> disabled and Export Timing --> disabled
(Console>configuration>setting) |
Hi, thanks for the reply. The graph export feature I was actually already disabled since I was not using it. But the main problem that causing the CPU usage to 100% is when viewing the graph. |
|
| Back to top |
|
|
BSOD2600
Cacti Moderator
Joined: 08 May 2004
Posts: 6763
Location: WA, USA
|
| Posted: Wed Sep 03, 2008 3:13 pm Post subject: Re: CPU usage issue & Error |
|
|
| tripleK wrote: | | But whenever I was viewing the graph, the CPU usage is 100% and the process that causing it to 100% is w3wp.exe. What is the actual process that was running in the background? Is there a way to reduce the processor usage while viewing the graph. |
IIS spawns cmd.exe, which in tern launches rrdtool.exe with the commandline arguments it was passed. rrdtool.exe is the actual process (one per graph I believe) which creates the graph you see from the rrd file(s). rrdtool does not have a processor limiting option built into it (nor does any typical application I know of). Short of segmenting your graphs/devices into logical trees/groupings so it doesn't have to graph so many things at once, I don't think there is any other way to lessen the CPU impact.
Odd the IIS service itself is using all the CPU. You do actually get functional graphs? hmm, wonder if its actually the php5isapi.dll module which is using the cpu in some thread...
| tripleK wrote: | | I did a snmpwalk on those DS and the snmpwalk return no results. How can I do reindex the OID? Will it cause my previouse data lost? |
Well if a snmpwalk doesn't return data for that OID, then neither will Cacti. Try removing a number at the end of what you snmpwalk to see if that'll return data instead. You are sure that snmp is functional on that device, right?
No reindexing will not cause you to loose the data, it simply tells cacti that the OID that it was tracking for something (lets say a NIC) has changed from X to Y.
All the data Cacti collects/graphs is stored in the rrd files. |
|
| Back to top |
|
|
tripleK
Joined: 02 Sep 2008
Posts: 4
|
| Posted: Mon Sep 08, 2008 2:57 pm Post subject: Re: CPU usage issue & Error |
|
|
| BSOD2600 wrote: | | IIS spawns cmd.exe, which in tern launches rrdtool.exe with the commandline arguments it was passed. rrdtool.exe is the actual process (one per graph I believe) which creates the graph you see from the rrd file(s). rrdtool does not have a processor limiting option built into it (nor does any typical application I know of). Short of segmenting your graphs/devices into logical trees/groupings so it doesn't have to graph so many things at once, I don't think there is any other way to lessen the CPU impact. |
Thanks for the explanation. I would think of dividing the graphs if there is the only way. Btw, I saw a boost plugin, and would like to seek some comment from some experts. 
My current cacti server is updating graphs every minute and about 3 engineers is required to keep monitoring the graphs concurrently. Will the boost plugin helps in this way?
| BSOD2600 wrote: |
Well if a snmpwalk doesn't return data for that OID, then neither will Cacti. Try removing a number at the end of what you snmpwalk to see if that'll return data instead. You are sure that snmp is functional on that device, right?
No reindexing will not cause you to loose the data, it simply tells cacti that the OID that it was tracking for something (lets say a NIC) has changed from X to Y.
All the data Cacti collects/graphs is stored in the rrd files. |
Well, I am lost on reindexing, may be some step by step guidance would help, any link?.  |
|
| Back to top |
|
|
BSOD2600
Cacti Moderator
Joined: 08 May 2004
Posts: 6763
Location: WA, USA
|
| Posted: Mon Sep 08, 2008 9:56 pm Post subject: |
|
|
Yes Boost will help since it'll cache the generated images, although since you have it refreshing every minute, maybe it won't help that much... There is a thread in the annoucement forum where users with large cacti installations have posted their metrics/configs. Should read through that to get an idea how to tweak cacti.
As for reindexing, have you looked through the documentation site (link above)? The basic way of reindexing a single host for a single data query is to click the green circle next to the query (in a device). |
|
| Back to top |
|
|
|
FAQ
Search
Memberlist
Register
Profile
Log in to check your private messages
Log in
