graphs not constantly drawn (gaps), however values OK

[parcival]

Hi all,
i have here zabbix v1.8.4 on Centos 5.5, 4core with 4GB RAM.
My Problem is that grahs not constantly drawn, however values OK.


graphs


values


Possibly an idea? Thanks for each assistance.

parcival

[parcival]

Can someone say, in which table the values for Graphes are stored?

thx

[nelsonab]

First off, do you really need to monitor your processor load every 5 seconds? This is the template default, but you may wish to go in an edit the template to be something a little more sane like 30 seconds or 1 minute.

Next if you have any gap in data longer than about twice your interval (i've have to look at the code to know exactly) there will be a gap drawn in your graph. For instance if you are monitoring something every 5 seconds, and you have data points as 0sec, 5, 10, 40, 45 and 50. You will see the graph drawn correctly between 0 and 10 seconds, followed by a gap to 40 seconds, then a graph to 50 seconds.

What is causing your gap. It can be a number of things. Most often there is a resource starvation occurring somewhere. The agent may not be getting enough CPU time to service the requests, or there may be network glitches, or the server may be overloaded to send the requests in time.

[parcival]

nelsonab, you have right with all things.
However i must explain the whole more exactly.
I have two zabbix installations (140 values/sec), one with FreeBSD and other with Centos. These are identical both up to two things, OS and DB Backend.
At FreeBSD works percona mysql Server without any problems.
Similarly i work with Centos and toku DB (fractal tree).

http://tokutek.com/

Toku DB would be an ideal replacement for innodb.
But i have here this "gap" graphs problem and do not understand why.
By feel i believe "House Keeping" is this problem. After "House Keeping" values in "history table" are missing and i get graphs gaps.

Maby house keeping have problems with this features from toku db.

Hot Schema Changes let you rapidly adapt to changing market conditions even on terabyte or larger databases: In a first for MySQL, TokuDB enables dynamic and rapid changes to database indexes and columns, allowing customers to remain agile with Big Data. Now index creation and column addition / deletion can run concurrently with read and write operations

No performance degradation, maintenance headaches: immunity to database aging means no more query slowdowns and no more dump/reload.

It would be super if zabbix with toku would work.

Main features:
very fast, online optimize, inbuild compress 30GB -> 7GB

Stefan

[nelsonab]

Before looking to replace your database engine, how much effort have you put into performance tuning your database server?

[parcival]

Thanks for your fast answer.
The zabbix instance with toku DB accurately with the documentation of toku DB one optimized. I see also no problem to performance with OS (4Core 4GB RAM). This problem must have another cause, however i do not understand current which.

[MrKen]

Hi all,
i have here zabbix v1.8.4 on Centos 5.5, 4core with 4GB RAM.
My Problem is that grahs not constantly drawn, however values OK.


graphs


values


Possibly an idea? Thanks for each assistance.

parcival

In the above, you have a graph from 15:41 to 16:18, but the values you show are for 16:18 to 16:20. These values are not from the graph period that you show.

What you should be showing is the values for 16:03:30 to 16:18. Are there any values for the 'gaps' in the graph? My guess is that there are no values for that period.

Do you monitor the zabbix_queue? Are there a lot of items in the queue during the periods of no data?

I must agree with nelsonab, updating the values every 5 seconds is ridiculous! Change this to every 30 seconds or 1 minute.

140 values/sec is not a lot, however 4 GB of RAM might be struggling. Perhaps you could consider increasing that to 8 GB.

MrKen