Zabbix configuration

[jrprado]

I would like your help to configure zabbix correctly with my environment. These items listed below are those who see by default. For what value do I change the items? Is there any calculation? What do these items affect the system or the zabbix?

zabbix_config.conf
# StartPollers=5
# StartIPMIPollers=0
# StartTrappers=5
# StartPingers=1
# StartDiscoverers=1
# StartHTTPPollers=5
# CacheSize=8M
# CacheUpdateFrequency=60
# StartDBSyncers=4
# HistoryCacheSize=8M
# TrendCacheSize=4M
# HistoryTextCacheSize=16M

Status of Zabbix


Uploaded with ImageShack.us

My hardware config:
AMD Opteron(tm) Processor 852 2,5GHz
Mem : 4GB
Database in another machine: Oracle 10g

These are some questions I have.

I thank the attention.

[jrprado]

Zabbix Server - Configuration

Could someone help me?

[untergeek]

Configuration parameters are very subjective.

How many "hosts" are you monitoring, or plan to monitor?

What's the performance requirement in "items per second" from the dashboard?

What growth will you have? Will you be using discovery? Can your agents/items be monitored using Active rather than passive monitoring? Will you be using Web monitoring?

Without knowing some or all of these, it would be very hard to give you an idea of what you need to change. You haven't even specified what OS you're using. That can make a difference.

I can tell you, as an Oracle backend user with 432 hosts being monitored and 331 values per second anticipated performance required, doing discovery and using Active monitoring where possible we have the following (only provided as guidelines):

StartPollers=20
StartIPMIPollers=0
StartTrappers=100
StartPingers=10
StartDiscoverers=40
StartHTTPPollers=1
CacheSize=48M
CacheUpdateFrequency=60
StartDBSyncers=64
HistoryCacheSize=16M
TrendCacheSize=8M
HistoryTextCacheSize=16M

[jrprado]

Hi untergeek, thanks for reply.

I gave the information you requested on the image above.

Following data:
Number of hosts (monitored/not monitored/templates) 612 - 561/16/35
Number of items (monitored/disabled/not supported) 30735 - 18663/12063/9
Number of triggers (enabled/disabled)[problem/unknown/ok] 6013 - 2249 / 3764 [28/195/2026]
Number of users (online) 36/2
Required server performance, new values per second 149.63

Most items are monitored via SNMP or agent. At the moment most of my existing servers and services configured in zabbix missing very little. I do not use discovery.

Already checking web, I monitor 14 sites.

Right now I'm even having problems with the queue of checks that are delaying much, several items in the queue over 10 minutes. Already changed the StartDBSyncers to 12 but without any improvement.

[EnigmA-X]

It's not easy, but to start I suggest to raise the number of 'StartTrappers' to 30 and see if that improves your situation. Within version 1.8 you can easily raise this value to 80 or even 100.

Also, you can check the performance of your database and monitor the zabbix cache (create an item, zabbix internal and use the keys zabbix[wcache,...,...] and zabbix[rcache,...,...]. Check http://zabbix.com/documentation/1.8/ for details on this items).

Keep us posted

[untergeek]

Based on the things you are specifying, I would suggest that you need to increase your Pollers.

All SNMP checks and passive agent checks are initiated by the pollers. If you don't have enough pollers you could have the problem you're seeing.

I recommend trying these adjustments:

# StartPollers=20
# StartTrappers=50
# StartPingers=3
# StartDBSyncers=24

Give these a try and see what your queue looks like after that. I'd like to hear back.

[jrprado]

Personally, I made some changes requested in zabbix. Is configured as follows now:

StartPollers=10
StartTrappers=10
StartPingers=10
StartDiscoverers=1
StartHTTPPollers=1
HousekeepingFrequency=1
CacheSize=32M
CacheUpdateFrequency=60
StartDBSyncers=12
HistoryCacheSize=16M
TrendCacheSize=8M
HistoryTextCacheSize=16M

I realized that there was a big improvement in performance of zabbix, as you can see in the queues.



It also follows the graph with data collected by the zabbix wcache and rcache. I saw that the text is with 0% free.



Should I change the HistoryTextCacheSize?

[alixen]

Hi,

Personally, I made some changes requested in zabbix. Is configured as follows now:

StartPollers=10
StartTrappers=10
StartPingers=10
StartDiscoverers=1
StartHTTPPollers=1
HousekeepingFrequency=1
CacheSize=32M
CacheUpdateFrequency=60
StartDBSyncers=12
HistoryCacheSize=16M
TrendCacheSize=8M
HistoryTextCacheSize=16M

I realized that there was a big improvement in performance of zabbix, as you can see in the queues.

You should consider raising StartPollers for 2 reasons:

1. you still have a lot of items late.
2. in case a lot of hosts are unreachable (network outage), each poller will wait until timeout expires before switching to another host. It will affect all hosts even those that are still reachable.

Should I change the HistoryTextCacheSize?

Yes, of course.
Make sure that your server has still some memory available since allocating to much cache may force it to swap.

Regards,
Alixen

[jrprado]

When I changed the HistoryTextCacheSize = 32M not started and the zabbix appeared the following message:

1546:20101116:140445.522 Cannot allocate shared memory of size 33554695 [Invalid argument]
1546:20101116:140445.522 Could not allocate shared memory for history text cache.

With this decreases the value. Is currently set as follows:

StartPollers = 20
StartTrappers = 10
StartPingers = 10
StartDiscoverers = 1
StartHTTPPollers = 1
HousekeepingFrequency = 1
CacheSize = 32M
CacheUpdateFrequency = 60
StartDBSyncers = 12
HistoryCacheSize = 16M
TrendCacheSize = 8M
HistoryTextCacheSize = 24M

Anyone know why this error occurred? Since I have more than 1 GB of memory free.

Soon as I inform the zabbix is behaving with such changes.

[EnigmA-X]

Which version of Zabbix are you running (on your server)?
Which operating system?

[jrprado]

SO

Linux debian 2.6.32-5-686 #1 SMP Sat Oct 30 22:47:19 UTC 2010 i686 GNU/Linux

Zabbix Server

 16295:20101119:174855.968 Starting Zabbix Server. Zabbix 1.8.3 (revision 13928).
 16295:20101119:174855.968 **** Enabled features ****
 16295:20101119:174855.968 SNMP monitoring:       YES
 16295:20101119:174855.968 IPMI monitoring:       YES
 16295:20101119:174855.968 WEB monitoring:        YES
 16295:20101119:174855.968 Jabber notifications:  YES
 16295:20101119:174855.969 ODBC:                   NO
 16295:20101119:174855.969 SSH2 support:          YES
 16295:20101119:174855.969 IPv6 support:           NO
 16295:20101119:174855.969 **************************

[subba5678]

Hi All,

I was also facing some issues with the queues . But this is related to Graph queue , i am not getting the latest data Graph only for the Host Status , and for remaining all the services i am getting the graph as well as Latest data ,For Host Staus i am getting the error as " No Data" in the graph . It happend for me in the test server as well as Production server , so in the test server i disabled the zabbix Agent and enabled it after that i restarted the zabbix agent , its being worked , but in the production environment i cant stop the zabbix Agent , is there any way to fix this problem without restarting the agent . I found that some thing is wrong on the Queue , Kindly please help me out . Please find the attachment of the screen shot as reference and the template which i am using as well as My Zabbix Config file as well . Kindly Please help me out.

{Local Host:status.min(300)}=2

Note: In both (Production as well as Test) environment facing the problem only with the agent

Thanks,
Subbu

[subba5678]

Hi,

I updated the zabbix_server.conf with the changes posted in the forum , but still i am not able to find last data . Please help me out.

Thanks,
Subbu