Gaps in bandwidth graphing

[TBT]

Back in January 2020 we experienced gaps in bandwidth graphing. We resolved the issue by making the following config adjustments.

A. Adjusted load_threshold = 10.0 to load_threshold = 40.0 in npcd.cfg
B. Adjusted Forks: 4 to Forks: 16 in mrtg.cfg

Adjusting the number of forks seemed to resolve our issue, however, we're experiencing the issue once again. Below is an example.

graph.png

Also some output of grep 'MAX load reached' /usr/local/nagios/var/npcd.log

Code: Select all

[01-12-2020 01:02:03] NPCD: WARN: MAX load reached: load 16.870000/10.000000 at i=1
[01-12-2020 01:02:18] NPCD: WARN: MAX load reached: load 15.880000/10.000000 at i=1
[01-12-2020 01:02:33] NPCD: WARN: MAX load reached: load 14.620000/10.000000 at i=1
[01-12-2020 01:02:48] NPCD: WARN: MAX load reached: load 13.160000/10.000000 at i=1
[01-12-2020 01:03:03] NPCD: WARN: MAX load reached: load 11.940000/10.000000 at i=1
[01-12-2020 01:03:18] NPCD: WARN: MAX load reached: load 11.480000/10.000000 at i=1
[01-12-2020 01:03:33] NPCD: WARN: MAX load reached: load 10.030000/10.000000 at i=1
[04-19-2020 01:01:45] NPCD: WARN: MAX load reached: load 40.030000/40.000000 at i=0

1. Would you recommend adjusting the fork value to 20, will it drastically affect performance of the server (CPU)?
2. What is the upper limitation on the fork value?
3. Do you have any further suggestions?

[tgriep]

1. Yes, you could increase the number of forks and is should not effect the performance very much.
2. I do think there is an upper limit for the value but going very large you will have to increase the open file limit on the server.
3. Further suggestions is to go through the MRTG config files in the /etc/mrtg/conf.d folder and remove the configs of devices that are not is service anymore. That will speed up the process as well.

[TBT]

1. Yes, you could increase the number of forks and is should not effect the performance very much.
2. I do think there is an upper limit for the value but going very large you will have to increase the open file limit on the server.
3. Further suggestions is to go through the MRTG config files in the /etc/mrtg/conf.d folder and remove the configs of devices that are not is service anymore. That will speed up the process as well.

1. Okay. I'll start with 20 and perhaps adjust to 25.
2a. What do you think that limit is 100, 1000?
2b. CentOS6 open file limit /proc/sys/fs/file-max default is 1003464, what are you suggesting?
3. Yes, I'm in the process of doing that too. MRTG clean-up has been a long outstanding issue and feature request since 2011, can you provide an update on the status?

Request: https://support.nagios.com/forum/viewto ... ilit=+mrtg

[tgriep]

I would guess that if you increase the number of open files, network connections, memory, you could keep increasing the forks. I did not find any information on the max but every system is different and would have to be monitored to see if it hits a limit.

There has been no update on the removal of the MRTG files feature request.