RRAs are not updated after host recovers

[NitrousOxyde]

Comrades,

After some host becomes down and recovers after some time automatically, its graphs are no longer being updated. I can, that relevant RRAs are not being updated at all as if spine does not "see" them and they dont exist on server

I've upgraded cacti and spine to the latest version(1.1.16), and without success - old graphs of hosts, which did not become down, are updating; graphs of hosts, which went down and recovered - not

Interestingly, when I navigate to devices, problematic devices are shown as "UP", but their uptime is "0d:0h:0m"

RRA ownership is checked and is cacti:cacti, all RRAs have permission 777

Any ideas why spine does not update relevant RRAs are welcome

[Osiris]

Data collection done or permissions.

[NitrousOxyde]

Nope, doesn't seem to be permission or owner issue. As I stated before, its mod is 777(was originally 644, changed for being sure), and ownership is cacti:apache

There is virtually no difference between RRDs, that are being updated and ones, that are not

[cigamit]

Did you check the Cacti log and see if those devices are properly returning Data to Cacti (maybe run one of the hosts in debug mode)?

If all else fails, you can try my debug plugin and try to see exactly where the process is breaking down.
https://github.com/cigamit/plugin_debug

[NitrousOxyde]

I've tried to install plugin, but cacti does not see it in the list of plugins. All files are transferred to debug folder in plugins folder

ls -l /usr/share/cacti/plugins/debug/
total 56
-rw-r--r--. 1 root root 12966 Aug 19 11:36 debug.php
-rw-r--r--. 1 root root 574 Aug 19 11:36 debug.sql
-rw-r--r--. 1 root root 226 Aug 19 11:37 INFO
-rw-r--r--. 1 root root 18046 Aug 19 11:37 LICENSE
-rw-r--r--. 1 root root 1089 Aug 19 11:37 README.md
-rw-r--r--. 1 root root 7120 Aug 19 11:37 setup.php

[NitrousOxyde]

Issue still continues - RRAs of some devices are being updated while others stop updating, then they change roles

[NitrousOxyde]

Hi comrades,

After increasing number of maximum poller processes to really high valuse (800 in my case) solved the problem - now all graphs are updated in time

"SYSTEM STATS: Time:20.8783 Method:spine Processes:800 Threads:1 Hosts:181 HostsPerProcess:1 DataSources:11343 RRDsProcessed:6187 "

Spine is used as poller, settings screenshot is attached

[cigamit]

Wow, ya changing it to something that high is something I would not suggest at all.
If using spine, your processes should be set to the same as the number of CPU Cores. The threads can then be changed to ~12 or higher as needed.

[Grandalf00]

What kind of interface/oid could you graphs? It seems like that oid (exactly index) is dynamic, for example pppoe on Mikrotik. Try re-index method set to "Verify all".

[NitrousOxyde]

What kind of interface/oid could you graphs? It seems like that oid (exactly index) is dynamic, for example pppoe on Mikrotik. Try re-index method set to "Verify all".

I am graphing interface statistics (input/output values in bps), as well as CPU and memory usage of equipment. Network is quite large, with a lot of routers/switches

Wow, ya changing it to something that high is something I would not suggest at all.
If using spine, your processes should be set to the same as the number of CPU Cores. The threads can then be changed to ~12 or higher as needed.

Well, this is the only possibility to be able to poll all hosts and update all RRAs. Increasing number of processes for spine affects negatively cacti's performance. BTW, spine is used for polling, not cmd.php