Process for cacti (1 poller)

[glamas]

Hello try to look for information about it and the truth that the search in the forum does not help much.
Here is the question as I can know in the cacti that the processes or the equipment is running small and I need to add another poller.
I have tried with gandalf template but I want to know if I am well located or there is another way.
From now on thank you

[TheWitness]

I did not quite understand your question. What version of cacti?

[glamas]

1.2.22

What I want to know is how to realize that the computer can not do all the queries and has loss of graphics due to hardware problems or configuration of the cacti

[TheWitness]

Use top or htop. Use more processes or threads if you have spine. There are article on this forum where people through the years have shown off their runtime stats. Just use that very long post as your guide post. Technical Support page is good too, but just for giving you the highlights. more Memory, cores, and NVMe/SSD is best practice these days. And if you get into the 10's of thousands of devices, having Remote Data Collection enabled.

[glamas]

I refer to this I think that I am under the processes of rrd at a time but I do not know if it is a matter of saturation of processes for example but I do not see anything strange.

[Rno]

Look into you cacti log , and you should look for this kind of line:

Code: Select all

13/02/2023 10:01:14 - SYSTEM STATS: [b]Time:12.2091[/b] Method:spine Processes:5 Threads:12 Hosts:5694 HostsPerProcess:1139 DataSources:18603 RRDsProcessed:0 

Time is suppose to be lower than your CRON pooling setting.

And if you have a problem of saturation you should find this entry on your log:

Code: Select all

13/02/2023 04:43:01 - POLLER: Poller[Main Poller] PID[31626] WARNING: There are 1 processes detected as overrunning a polling cycle, please investigate 
13/02/2023 04:43:00 - SYSTEM STATS: [b]Time:58.8971 [/b]Method:spine Processes:4 Threads:13 Hosts:655 HostsPerProcess:164 DataSources:9099 RRDsProcessed:6253
13/02/2023 04:43:00 - POLLER: Poller[Main Poller] PID[30787] Maximum runtime of 58 seconds exceeded. Exiting. 

This time it take 58.8971 seconds to process all RRD, a little to long for my 1 minute CRON.
On my side I have this little situation when some equipment take to long to answer a SNMP query

[TheWitness]

You need to enable boost, oh, and you better have a lot of cores on your cacti server with those settings, like at least 10 if not higher. Also, even with boost, if you count the size of all the RRDfiles, it's good to have that size plus 2-4GB More of server memory unless you are also using plugins that use a lot of MySQL tables like Syslog, MacTrack or Flowview. If using those plugins the database memory requirement can be high, so more physical memory.

If in a corp environment and being jammed into a VM, watch out for snapshots, which will cause you cacti server to lockup.

The VM admins will say it only takes a few seconds, but when your vdisk is a few TB, they get quiet real quick.

[glamas]

hello the witcher if I encounter a problem I have enabled in cron.d


2,7,12,17,22,27,32,37,42,47,52,57 * * * * root /usr/bin/php /usr/share/cacti/poller.php > /dev/null 2>/var/log/cacti/ poller-error.log
0,5,10,15,20,25,30,35,40,45,50,55 * * * * root /usr/local/spine/bin/spine/spine -V 4 > /dev/null 2>/var/log/cacti/po ller-error.log


When I disable the poller in cron.php it is not graphing me.

However on the cacti menu it says I'm running spine.