Gaps in graphs

[frankp]

Hello all,
I've added a new LINUX host to my Zabbix server.
I've choosen the default "Template_Linux" and deactivated a few of the items.
The CPU graphs are looking chopped like the one attached.
The zabbix_server.log does not contain realy useful information regarding this behavior. Anyway, I've attached the log as well.

Any ideas and explanations are welcome.

[Akao]

Same happens to me using snmp agent in some hosts.

Logs show timeouts and network errors.

[jeenam]

Try changing the following paramter in zabbix_server.conf:

### Option: Timeout
# Specifies how long we wait for agent, SNMP device or external check (in seconds).
#
# Mandatory: no
# Range: 1-30
Timeout=30


Sounds as though your client is not responding to the server query within the timeout period.

[frankp]

Sorry, but changing the timeout value to 30 did not change the situation.
Any further ideas?

[frankp]

Morning,
even if no one has a perfect solution, a view hints and tipps to come closer to a possible solution would be great.
We noticed the described behavior with Linux/Unix systems only. Windows systems seem not to be affected.
To get rid of these problems is very important for us.
Any help would be very much appreciated.
Thanks
Frank

[tchjts1]

Frank - I've had this issue off and on for the past 3 years across various versions of Zabbix.

On earlier versions of Zabbix, one solution that seemed to work for most of the issue was to change from using passive items to active items - Zabbix agent (Active). For this to work, you must have a line in your zabbix_agentd.conf file: Hostname=xxxxxxx where xxxxxx is an exact match to the name of the server you have in Zabbix frontend. If you make agent conf changes, restart the agent.

I would also be curious as to how you have your Server=xxxxx line defined in zabbix_agentd.conf - are you using IP or DNS entries on that line? We have found that if we use DNS entries and there is any problem with the host resolving DNS, it can cause problems. We had some hosts with bad nameserver entries in their resolv.conf and it totally messed up the majority of hosts graphing.

[frankp]

Morning tchjts1,

thanks for your reply. Changing from using passive items to active items seams to solve the problem.

It would be great if a developer could give a more detailed explanation and maybe an outlook when this will be solved for passive items also.

As additional information to tchjts1's suggestions:
We use IP addresses within the "Server" line in zabbix_agentd.conf. So DNS shouldn't be the cause of the problem.

Many thanks and a nice day
Regards
Frank

[Aly]

Generally such problem occurs when server performance is low. Mostly could be resolved by DB or/and ZBX Server parameter tuning.