I am having a issue with CPU utilization being high during the NC_NET checks on Windows 2003 SP2 machines. I am running NC_NET client 2.28 and have tried 4.4 on these systems with the same results. I am running Nagios server version 2.7. I have a screen shot but don't know how to get it posted at this time. For example I have a system that when the NC_NET.exe runs it is using 13% CPU and the services.exe is using 34 %. It doesn't seem to let up at all. Is there a way to throttle this back?
Thanks
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I can understand your issue,
the source code for that check is roughly the same since v1.X of NC_NEt.
It is always a ballance of Reliability, Speed, accuracy, and Load.
currently NC_NEt has no settings to tweek this usage,
It uses the Dot Net Framework to get a list of the Services,
Itterate throught them, and check their status,
the usage is due to Microsofts Dot Net interface to the services.
Actually I usually found Opening Service Manager in windows also typically induces a High load and wait time.
Can any of these be tested through Counters? of the Process category that may be the least intrusive on Host Load.
An Alternative interface to check Service state may be through WMI
this is also availible through Newer version of NC_NEt.
I recomend upgrading NC_NEt if availible, there several improvements included in it,
to security, stability and connectivity as well as new checks unavailible to v2.28 including RUNSCRIPTS, EVENTLOG_NEW, Servertime,
it has support for COMMANDFILES, that run Several NC_NET commands through a single check_nt call.-(this is more for fetching lots of performance data with little concern of the output string or status OR for converting the standard output to a different output stream like wrapping the results with echo commands)
TOny
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Thanks Tony... I noticed that the older versions of NC_NET use Dot Net 1.x and the new version uses Dot Net 2.x. Does Dot Net 2.x interface to the services work any better than 1.x? If not I will look into upgrading the NC_NET client to 4.4 and check out WMI calls.
I have one other issue with the a 2008 Windows 64bit system. I have the new NC_NET (4.4) client installed and need to change the default listen port from 1248. In the past with the 2.28 client all I did was change the port in the startup.cfg file and restart the service. When I try that with the new client I get "BAD Command for startup config:port 1984" in the Windows Application event log and the port stays at the default 1248. Is there another way to change the listen port with the 4.4 client?
Thanks
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Wanted to follow up with what the issue was on the high CPU utilization. In working with microsoft on another issue this was brought to their attention and they said that typically the high CPU on the services.exe is related to eventlog checking. I know that there is a issue with the older NC_NET client creating high load on eventlogs bigger that 5mb. So I looked and someone had increased the size to 16mb and the logs were actually that size. Dropping the log size back to 5mb solved the issue now the NC_NET.exe and services.exe conbined only get to 12% CPU only for a few seconds. Does the new 4.4 client using the EventLog_NEW solve the 5mb eventlog size limitation?
Thanks
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Providing you are looking for the Latest Results from the Event Log
Event_Log New MAy be the best source to get it.
And Log file size has almost NO impact on the check!!!
Please test it an post back your results in a new thread.
I only tested EventLog New for Completion Time,
an average 20s check should return alost instantly.
Also it was only recomended to do one eventlog check using the old event log check
The new version can have consecutive checks made against it.
THe input syntex is also new I hope it is easier to use,
It accepts Rages and comma seperated lists for the EventID
Regular expressions are implemented in the Event log check.
All RegExp options are configurablethrough the startup config
since Dot Net does not handle RegExp the same as Unix but can be
converted via the parameters.
I have to look into the Port# issue since your report is the first I heard of it, and that version has been out for a Year already.
Email me direct ANd I can give you a 64bit copy of NC_Net using Dot Net 3.5 if your interested.
I have not publicly released it since the install package and documentaion are not yet completed.
Tony
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
After re-reading your post (as well as testing locally)
The Startup.cfg uses the TAB between the parameter and the value.
The Event Log entry is the line that had the error.
ALso the startup.cfg does not remove Emptyspace on lines
Tony
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I am having a issue with CPU utilization being high during the NC_NET checks on Windows 2003 SP2 machines. I am running NC_NET client 2.28 and have tried 4.4 on these systems with the same results. I am running Nagios server version 2.7. I have a screen shot but don't know how to get it posted at this time. For example I have a system that when the NC_NET.exe runs it is using 13% CPU and the services.exe is using 34 %. It doesn't seem to let up at all. Is there a way to throttle this back?
Thanks
I can understand your issue,
the source code for that check is roughly the same since v1.X of NC_NEt.
It is always a ballance of Reliability, Speed, accuracy, and Load.
currently NC_NEt has no settings to tweek this usage,
It uses the Dot Net Framework to get a list of the Services,
Itterate throught them, and check their status,
the usage is due to Microsofts Dot Net interface to the services.
Actually I usually found Opening Service Manager in windows also typically induces a High load and wait time.
Can any of these be tested through Counters? of the Process category that may be the least intrusive on Host Load.
An Alternative interface to check Service state may be through WMI
this is also availible through Newer version of NC_NEt.
I recomend upgrading NC_NEt if availible, there several improvements included in it,
to security, stability and connectivity as well as new checks unavailible to v2.28 including RUNSCRIPTS, EVENTLOG_NEW, Servertime,
it has support for COMMANDFILES, that run Several NC_NET commands through a single check_nt call.-(this is more for fetching lots of performance data with little concern of the output string or status OR for converting the standard output to a different output stream like wrapping the results with echo commands)
TOny
Thanks Tony... I noticed that the older versions of NC_NET use Dot Net 1.x and the new version uses Dot Net 2.x. Does Dot Net 2.x interface to the services work any better than 1.x? If not I will look into upgrading the NC_NET client to 4.4 and check out WMI calls.
I have one other issue with the a 2008 Windows 64bit system. I have the new NC_NET (4.4) client installed and need to change the default listen port from 1248. In the past with the 2.28 client all I did was change the port in the startup.cfg file and restart the service. When I try that with the new client I get "BAD Command for startup config:port 1984" in the Windows Application event log and the port stays at the default 1248. Is there another way to change the listen port with the 4.4 client?
Thanks
Tony
Wanted to follow up with what the issue was on the high CPU utilization. In working with microsoft on another issue this was brought to their attention and they said that typically the high CPU on the services.exe is related to eventlog checking. I know that there is a issue with the older NC_NET client creating high load on eventlogs bigger that 5mb. So I looked and someone had increased the size to 16mb and the logs were actually that size. Dropping the log size back to 5mb solved the issue now the NC_NET.exe and services.exe conbined only get to 12% CPU only for a few seconds. Does the new 4.4 client using the EventLog_NEW solve the 5mb eventlog size limitation?
Thanks
the event Log New is great.
Providing you are looking for the Latest Results from the Event Log
Event_Log New MAy be the best source to get it.
And Log file size has almost NO impact on the check!!!
Please test it an post back your results in a new thread.
I only tested EventLog New for Completion Time,
an average 20s check should return alost instantly.
Also it was only recomended to do one eventlog check using the old event log check
The new version can have consecutive checks made against it.
THe input syntex is also new I hope it is easier to use,
It accepts Rages and comma seperated lists for the EventID
Regular expressions are implemented in the Event log check.
All RegExp options are configurablethrough the startup config
since Dot Net does not handle RegExp the same as Unix but can be
converted via the parameters.
I have to look into the Port# issue since your report is the first I heard of it, and that version has been out for a Year already.
Email me direct ANd I can give you a 64bit copy of NC_Net using Dot Net 3.5 if your interested.
I have not publicly released it since the install package and documentaion are not yet completed.
Tony
After re-reading your post (as well as testing locally)
The Startup.cfg uses the TAB between the parameter and the value.
The Event Log entry is the line that had the error.
ALso the startup.cfg does not remove Emptyspace on lines
Tony
Yep that was my bad it is a tab. Changed and it works fine.
Thanks....