Zabbix agent 1.8.2 reports disk space too high (wrong units by an order of magnitude)

[L2G]

I've recently built Zabbix 1.8.2 both as a server and an agent. They work great except for one hitch.

For the machine where I built the agent, disk space is reported too high on the web UI by an order of magnitude. In other words, megabytes are given as gigabytes, gigabytes as terabytes, etc. Meanwhile, for machines still running older agents, the units of disk space are still correct.

This is not a show-stopper, as the agent's values are still useful if we stick to percentages instead of absolute values. But now I'm afraid to upgrade the agent on any of our other machines for fear of messing up their histories.

For what it's worth, the machine with the agent is running an older Linux kernel (maybe 2.6.17). Please don't ask.

Is this a known bug and/or has anyone else come across it?

[jjf]

i'm using zabbix 1.8.1 and have the same issue. the reported value is off by an order of magnitude as well (1024x)

[simonc]

http://www.zabbix.com/documentation/...tion#upgrading

zabbix 1.8.x ->

Key vfs.fs.size returns data in bytes for all operating systems now.
Key vfs.fs.size now takes into account reserved diskspace for root user.

So for a new zabbix agent 1.8, you should reajust your 'disk space' triggers.

[L2G]

Oho, now I see

Now I see the gotcha that got me. My Zabbix already has a bunch of defined "items" for disk space. And while I did understand that space was now being reported in bytes instead of kilobytes, I did not remember that all of my existing disk-space-related items had a multiplier of 1024 defined, so did not realize that it would be my responsibility to fix those.

Well... yecch. :-) I guess that means whenever I upgrade an agent, I'll need to put that host into maintenance mode in Zabbix so it won't record erroneous values in the history, then fix the appropriate items. Oh, or a little more convenient: I could rename all of my current templates, then clone them and fix the clones; that way I know which ones are for version-1.8 values and which are for pre-version-1.8 values, and eventually phase out the latter.

[sean]

Hi,

I've just had this too..
Currently I've a template for all my Linux systems, does this mean I should then create new template, without the multiplier for all Linux systems with v1.8 clients?

[jjf]

yes. it would be advisable to have a "clean" template for v1.8, which takes into account the changes made from v1.6

[sean]

The upgrade from v1.6 to v1.8 went well a few weeks back, just noticed these problems now.

I tried to clone my Linux template, but get a "circular reference".

Then I exported it to XML, did a search and replace to rename the temple, and reimported.
Now it complains about undefined index ymin/max_itemid after adding the trigger for Low diskspace, and before adding a graph of System Load.
Its also noted here, but no follow-up:

https://support.zabbix.com/browse/ZBX-2330

So the question if how I can "clean" my template, or find a clean one.
The std Zabbix one for Linux is full of non Linux essential stuff like apache/mysql and so, if I remember correctly..

[jjf]

is your template linked with another template?

also can you attach your exported xml?

[sean]

Well thats very kind of you, the template is attached.

Yes there is one dependency, it links to "Template_SIS_ping". Which already exists and works fine (just a small template with to icmp items).