Graph Gap

[tchjts1]

Take a look at the bottom of this post regarding graphs for Zabbix internal process performance: https://www.zabbix.com/forum/showthread.php?t=41219

This will help you identify more specifically what needs to be adjusted in your Zabbix conf.

[Denis Bond]

Hello.
I still have an issue.
I checked the load of the zabbix server everything is fine. It's not overloaded and have plenty of CPU and memory resources.
I wonder if anybody managed to resolve the issue?

[aib]

Sorry for repeating the same questions: Did you check Zabbix Data Gathering Process Busy graph and Zabbix Internal Process Busy graph?

Sometimes you have CPU load around 0% (because you don't have enough processes) and in the same time one Data Gathering process can use all CPU (because you have only 5 ICMP poller by default and 100 devices, configured by yourself only for ping)

Check, please, that Zabbix Processes graphs

[dirckcopeland]

Graph Gap

Denis,
I am seeing the same thing with my Zabbix server. It seems to be only on the SNMP queries and on my system, the gaps are at pretty consisant intervals (around 5 minutes), and then data starts coming through again. I run tcpdump on the zabbix server and can see that the traffic stops and coincides with exactly as Zabbix is displaying it on the graph.

Also wanted to mention that when viewing the graph, I select the pulldown to view the data as values and sure enough, there is no data with a timestamp where there are gaps in the data on the graph.

Another oddity is when the data stops being ttransmitted (by viewing with tcpdump) I can still ping the system so it's not like there is a connectivity issues with being able to reach the system over the network. Also there are no firewalls in between or installed on either of these systems.

[Denis Bond]

Resolved

Hello.
I may not resolve it in a proper way but nevertheless....
I installed phpmyadmin and logged into the database and TRUNCATE history;
TRUNCATE history_uint; TRUNCATE trends; and TRUNCATE trends_uint;

and set the even recurring task to Truncate those tables monthly.

try it. But it may be not a good solution for you.

[ingus.vilnis]

Hi,

TRUNCATING those tables simply deletes all collected historical values. This has nothing to do with your server performance unless you are out of disk space.

In case of SNMP check status of bulk requests, are they enables or not and whether your devices support them, plus check Timeouts.

Best Regards,
Ingus

[dirckcopeland]

Graph gap

Ingus,
I don't believe I am using bulk requests, when I look at the traffic using tcpdump there is only one OID in each packet, not multiple as you would expect to see if your using bulk request.

Never the less, how do I check to see if bulk requests are enabled on the Zabbix server. A quick search in the Zabbix manual does not return anything about bulk requests.
Thanks much,
Dirck

[ingus.vilnis]

Hi Dirck,

Which version of Zabbix server are you running?
Checking bulk requests depends of version.

If you have 2.4.x then the setting can be selected for each host at host configuration card in frontend.

If you are using 2.2.x then till 2.2.2 there were no bulk requests at all.
They were introduced in 2.2.3.
In 2.2.7. a config setting in zabbix_server.conf was added EnableSNMPBulkRequests=1

Best Regards,
Ingus

[dirckcopeland]

Graph gap

Ingus,
I am running 2.2.2 so that would explain why I'm not seeing anything about bulk requests.

What is really strange is I set the latest data history to values to see what data is being stored, and this item has an update interval of 60 seconds, but as you can see in this attached screenshot, it shows values for 60 seconds and then there will be gaps of 8 to 12 minutes before data is receivied again.

thanks,

Dirck

[tchjts1]

Dirck - Are you still running with the stock settings in your zabbix_server.conf?

Can you take a look at this post https://www.zabbix.com/forum/showthread.php?t=47781

I would be curious to see what your Zabbix server performance is in those 3 graphs at a 1 day view.

[dirckcopeland]

Graph Gap

TCH...
Here are my setting in the zabbix_server.conf file. I do not see the "new internal Zabbix checks " that is referenced in the link you provided. Perhaps I need to import the "Template App Zabbix Server", I do not have that.

DBName=zabbix
DBUser=zabbix
StartPollers=30
JavaGateway=127.0.0.1
JavaGatewayPort=10052
StartJavaPollers=5
StartSNMPTrapper=1
SNMPTrapperFile=/tmp/zabbix_traps.tmp
CacheSize=128M
HistoryCacheSize=512M
TrendCacheSize=128M
Timeout=30
UnreachablePeriod=60
AlertScriptsPath=/etc/zabbix/alert.d/
ExternalScripts=/etc/zabbix/externalscripts
FpingLocation=/usr/bin/fping
Fping6Location=/usr/bin/fping6
LogSlowQueries=5000
ProxyConfigFrequency=600
StartVMwareCollectors = 1
VMwareCacheSize = 8M
VMwareFrequency = 60

[tchjts1]

Dirck - It is impossible to tell if your settings need adjusted without seeing the graphs mentioned in that post I linked to. I would be dumbfounded if they are not including Template App Zabbix Server with the 2.2.x distribution.

If you don't have it, absolutely go to this link download, import and attach it to your Zabbix Server: https://www.zabbix.org/wiki/Zabbix_T...s#Zabbix_2.2.0

It is under the "2.2.0" heading and is the 15th template down on that list.

But before you do that, look at the items that are listed under your Zabbix server. You do not see these listed? If not, then do as suggested above.

[dirckcopeland]

Graph Gap

tchjts1,
The templates were not included but I imported and applied them to the zabbix server. Here are the graphs for the system experiencing gaps in the data.

[tchjts1]

Your set up appears to be running pretty smoothly. If it were me, I would increment my ICMP pollers a bit and get that down in the 20% (or less) range.

You are missing an important graph though, refer to my link again and show your graph for your cache usage.

[dirckcopeland]

Graph Gap

tchjts1,
I was about out of space on my uploads so I had to remove one to attach this one. I'll attempt to post the latest graph on cache usage. By increasing the ICMP pollers, are you refering to the "StartPollers=30" setting?