zabbix_sender data randomly not arriving (Z v1.8.2)

[fascinatedcow]

Hi,

I have a few items that are of type "zabbix trapper". My hosts that have the items are monitored by a proxy. Every 24 hours I send (from a script) the data for the item to my proxy and my node (just in case I'm to move the host over to the node for some reason). I can see that the item is sent to the proxy and the node:

Info from server: "Processed 1 Failed 0 Total 1 Seconds spent 0.000081"
sent: 1; skipped: 0; total: 1
Info from server: "Processed 0 Failed 1 Total 1 Seconds spent 0.000041"
sent: 1; skipped: 0; total: 1

The first one gets to the proxy, which is what I would expect, but most of the time this data never actually arrives and so my nodata trigger fires. The strange thing is that this sometimes does work.

Also, I'm not convinced that this problem is specific to trapper data since many of my items that are only checked every hour or more end up "queued". This may be an entirely different problem, however.

I have tried increasing my StartTrappers to 50, but this has made no difference.

Anyone else seen similar behaviour?

Is there any way to see what's going on here?

Thanks,

Matt

[alebrz]

The same behavior is happening with me.
My infrastructure is different, but the symptom is the same.
My network hosts send traps to zabbix, but I don´t know why zabbix sender presents error like you show above.
To solve that problem I cloned the host. Then I set the same name and altered the name of the original device. I created a new zabbix trapper item to receive the trap.
After changes, when the trap was sent zabbix received the data.
Probably there is better way to solve it, but I dont know how ...

[James Wells]

Greetings,

The first one gets to the proxy, which is what I would expect, but most of the time this data never actually arrives and so my nodata trigger fires. The strange thing is that this sometimes does work.

Right off the bat, I suspect your DB is lagging. What you are describing is symptomatic of heavy DB loading. I am going to guess that your master DB is MySQL and you are using a single monolithic DB file for Zabbix.

If I am right, you can take some load off by converting to innodb_file_per_table, and take even more load off by tuning the DB a bit. Check out Mysql Database Grow (36 Gb) | How Optimize parameters for some ideas on how to tune the DB.

[fascinatedcow]

I thought I had replied to all of my threads I had open about seemingly (but in the end not-so) unrelated issues, but I guess I missed this one. For me the solution was DB-related, but it wasn't a tuning issue, it was because I was using the gcc-compiled version of mysql binaries. When I switched to icc every single one of my problems went away.