An Issue with Graphs for A Specific Device

[Mood977]

Hello All,

This is the first topic for me on the forum.

I need your support to identify the problem of my cacti, as there is a device its graphs are not drawing well as per below snap, while if I add the same device with the same parameters the new device's graphs working well.

I need to solve the existing device as it has around 2K graphs and I don't want to lose the history.
Also, I believe it's not a performance issue as the Cacti has many devices with the same amount of graphs and the other devices working normally.

Not Working graph.PNG (58.68 KiB) Viewed 658 times

Server's info:

Version.PNG (26.81 KiB) Viewed 658 times

Thanks in advance!
Sobhy

[eholz1]

Hello,

we might have to see the log file. But, gaps in graphs are usually due to changes in the poller, the collection profile, and how
often the poller runs (via cron). Gaps can occur if a change has been made in any of the variables above.

Usually this is not a permissions issue. You need to check poller process/threads, cron run time, and poller run time.
I used to see gaps when using poller period of "every 30 sec".
I set my cron to 5 minuts, and my poller to 5 minutes, I have 466 devices, RH7,PHP5.4, MariaDB 10. This setup runs 10 processes and 13 threads without issue.

I would consider changing poller to 5 min across the board.

eholz1

[Mood977]

Thanks eholz1 for your response.

Actually my cron interval is set to 5 min, please find the below snap.

Also, I need to mention if I added the same device as a new one the graphs will work normally without gaps, the thing is, I need the current one.

Poller.PNG (61.21 KiB) Viewed 650 times

Thanks,
Sobhy

[netniV]

You'll struggle with support for 0.8.8 as it's an old version now. You could put the device into debug mode to see what results you are getting back for the graph in question.

[Mood977]

Hello All,

Thanks for your feedback and support.

It has been recovered after I adjusted the SNMP Time Out for this device to 15000.

SNMP Time Out.PNG (3.16 KiB) Viewed 610 times

Regards,

[netniV]

Does that mean you've resolved it? If not, I would check that your poller is polling, there are no errors in the logs and that you have enough connections available to mysql for the number of threads/processes and users on your system.

[netniV]

I just saw the timeout post, which I missed when replying previously. I'd be very careful with having such a large timeout. If you do, make sure your retries are low, or your polling time is large. If you do not, you can stop other devices being polled because of one timeout.

[Mood977]

Hello netniV,

Plz could you share how to make sure the retries are low, is it an option from cacti side or from the device side and share how to do it.

Many thanks,

[netniV]

You have several settings, in the settings table for the defaults, against each device for individual settings. You can change on mass using the devices list and then ticking the ones you want to change, then use the Actions -> Change Device options.

[windsor]

Looks like you largely found the problem.

FWIW, it's a symptom that your device may be overloaded or buggy, if you're not reaching way across a WAN link for the information.

We have a couple of devices that do this when they get "cranky", basically the SNMP response time increases and cacti polling "gives up waiting." Coincidentally, SolarWinds starts reporting "excessive ICMP response time" on the same devices when they do this.

The hardware vendor tells us that the hardware de-prioritizes ICMP, but won't tell us how to disable that "feature." sigh.

We already have the new hardware in place to upgrade our "cranky devices."

[Osiris]

Solar winds basically has a threshold on the ping rtl and deviation from median or average. It's likely average as medians a bitch mathematically.