Graphs show scattered points

[abdulshafy]

Hello all,
I'm using Zabbix 1.8.3 on ubuntu server, and it's been fine for at least 10 days after its installation, but now it shows graphs as a scattered set of points when viewing a 1 hour history, however when viewing a 12 hours history they're displayed fine!

1 hour history>


12 hours history>


What can be the problem?
Any help is much appreciated.

[abdulshafy]

Still waiting for an answer.

[Aly]

Dots show that there is some data, but it comes much rare then specified by delay parameter.

[Zaniwoop]

I had the same problem.

You need to increase the number of pollers in the zabbix_server.conf file.

[abdulshafy]

Thank you very much, it seems to work now after increasing the number of pre-forked pingers to 10.
The question is how to determine the optimal number of pingers as the number of monitored devices grow?

[abdulshafy]

I have, so far, 70 monitored hosts, excluding Zabbix server, with a total of 427 items, all of which are simple checks, icmpping, icmppingsec and icmppingloss. The graphs got far better after I increased the pre-forked pingers to 100, but there's still some graphs that show dots! do I have to further increase the number of pingers?!!
Any hint would be much appreciated. Thanks in advance.

[abdulshafy]

Waiting for any further clarification...

[jonh]

I'm experiencing this same issue. I increased the number of pollers to 100 on both the server and proxy, but it did not resolve the problem. Any other suggestions?

It only happens at certain zoom levels on the graph. If I go out to a week it looks fine. But at 1 day, I just see individual points instead of the complete graph.

Edit: I should add this only started after upgrading to 1.8.3. CentOS 5.5 64-bit with MySQL 5.1.50 (also 64-bit).

[jonh]

I decided to open a support ticket, we'll see what they say.

https://support.zabbix.com/browse/ZBX-3034

[Zaniwoop]

The zabbix_server.conf file says " You shouldn't run more than 30 pollers normally."

a 100 sounds a bit excessive. I have mine set to 30.

I have 408 hosts, 11155 items and Required server performance, new values per second 207.04. I am not sure how these values compare?

[bashman]

I have my Pollers and Trappers set to 30, with 410 monitored hosts.

[abdulshafy]

I have set the number of pre-forked pingers to 30 but the problem still persists.

[bashman]

I have set the start pingers to 10, and I have 483 ping items:

mysql> select count(distinct hostid) from items where key_ like "%ing%";
+------------------------+
| count(distinct hostid) |
+------------------------+
|                    483 |
+------------------------+
1 row in set (0.01 sec)

Which would be the right number of start pingers?. Could it be 10 per 500 ping items?.

[tchjts1]

Zabbix 1.8.3
MySql DB

Number of monitored hosts: 1135
Number of monitored items: 69433
New values per second: 760.63

No dots on graphs.

My Zabbix Server conf:
StartPollers=35
StartTrappers=128
StartPingers=2

[bashman]

Zabbix 1.8.3
MySql DB

Number of monitored hosts: 1135
Number of monitored items: 69433
New values per second: 760.63

No dots on graphs.

My Zabbix Server conf:
StartPollers=35
StartTrappers=128
StartPingers=2

Which is your polling interval?, how is your zabbix queue?, do you have zabbix gui, zabbix server and db on the same host?, which is your io ratio?.