Zabbix graphs become dots

[jix]

Today my zabbix server start working strange

all of my graphs stop drawing regularly and become dots

my server average load and my database load seems to be ok

i don't know what happened and what is the cause of problem

also my pinger graphs works fine !!!

here is my configuration

start poller =40
number of item = 19000
new values per second 76.12

zabbix 1.8.10

any help will be appreciate

[jix]

!

I solved the problem
i think this will be happened when zabbix Queue become full of items that have more than 10 minute. . .

and also i find out that when one or some host become unreachable, my queue grows up and up until my zabbix server not working properly

maybe some kind of bug in zabbix server

any idea?

[jix]

!

The problem occurs sometimes ,
when one or several host become unreachable or with a lot of packet lost

my internal checks shows that my poller processes become 100% busy .

and my graphs stop drawing properly.

i increased my start poller from 18 to 40 but no success.

any experience or idea about this issue?

[ghoz]

Zabbix pollers are not event driven. Each process manages one and only one query at a time.

When a host is down, the poller don't get any answer and has to timeout (3secs by default) during which it cannot process any other item.

It's quite devastating when you have a whole network down as every item for every host in this network will be blocking your pollers ...

Say for example your average response is usualy under 500ms (quite slow) and you have just enought pollers to keep up with the queue.
Now if every poller get a item on a dead host, it will require 6 times your number of poller to kepp up with the increased delay... now you can watch your "queue" pile up ...

What are your options :

• monitor zabbix internal items like poller busy time etc and try disconnecting zabbix from the network, watch for the processes needs, adjust your configuration and increase the number of pollers accordingly
• set items as active, that way the pollers don't have to wait, the agent will send the data when and if it's ready...

[ghoz]

Oh and BTW your ping graph keep getting it's data because you have enough pingers process to keep up ...

[harmonica]

I,

We have been having the same problem but it only happens in high traffic ports in some of our routers. On the other ports the graphs are OK.



Our configuration

StartPollers=20
StartPingers=10
StartPingers=10
Number of item = 20018
New values per second 106.17

Zabbix Server 2.0.5

any help will be appreciate

[harmonica]

I,

The update interval for the item is 120 seconds. Changing to 60, 180, 240 or 300 seconds the dots are still visible.

Any help will be appreciate
Regards

[harmonica]

Hello,

I found solution for this issue.

Regards

[sevendollar]

The update interval for the item is 120 seconds. Changing to 60, 180, 240 or 300 seconds the dots are still visible.

We have been having the same problem but it only happens in high traffic ports in some of our routers. On the other ports the graphs are OK.

Hi harmonica, We happen to be running into exactly the same issue like yours, could you share the solution to the issue? many thanks.

[harmonica]

I,

The solution was start using snmp 64 bit counter.

https://support.zabbix.com/browse/ZBX-5339

http://www.cisco.com/en/US/tech/tk648/tk362/technologies_q_and_a_item09186a00800b69ac.shtml

[sevendollar]

It worked, appreciated for your help.