vfs.dev.read broken in Linux

[jeenam]

Server version = 1.8.2
Client agent version = 1.8.2
Arch = x86_64
OS = Centos 5.4

This has been on ongoing problem since 1.6.x...

The dev.vfs.read feature is flat out broken and hasn't worked since agent version 1.4.6. Creating the following agent check creates no graph data and latest data value is always '-'.

vfs.dev.read[sda1,sectors]
Type - numeric (tried float and unsigned)
Units - B/sec
Use Multiplier - Custom Multiplier
Custom Multiplier - 512
Store Delta - speed per second
Show Value - as is

Strangely enough, dev.vfs.write using the same parameters as above works great.


Why was this broken in the first place? Doesn't anyone else here like to see their disk read/write per second in MB?

[richlv]

both read and write work just fine for me. does it work with zabbix_get for you ?
can you try on a 32bit box, just in case it's architecture specific ?

[freakyflow7]

same issue

I'm having the same problems with these key. I would also like to have that kind of information available in ZABBIX! Thanks guys.

[freakyflow7]

Hi guys,

I found this link: http://www.muck.net/?p=19

This user re-created what vfs.dev.read/write would do.

Hopefully this helps while we wait to get some attention on this problem from the zabbix devs.

[jeenam]

Seen the muck.net posting before but the question remains...

Why is this broken and why has it been broken for so long? Will it ever be fixed? Seems like a trivial fix as it _used_ to work in 1.4.x.

Alexei, any chance of a fix?

[jeenam]

both read and write work just fine for me. does it work with zabbix_get for you ?
can you try on a 32bit box, just in case it's architecture specific ?

Richlv, using zabbix_get to pull the vfs.dev.read[device,sectors] value always seems to return the same value for me. It never changes, even after I KNOW my disk has been reading back data.

In addition, the graph for the item value doesn't show any value above 0. Is your item value actually creating graphs? Can you verify by posting the graph from your 1.8.x interface?

Thanks

[richlv]

Richlv, using zabbix_get to pull the vfs.dev.read[device,sectors] value always seems to return the same value for me. It never changes, even after I KNOW my disk has been reading back data.

In addition, the graph for the item value doesn't show any value above 0. Is your item value actually creating graphs? Can you verify by posting the graph from your 1.8.x interface?

Thanks

well, if zabbix_get works, retrieving by server and graphing should also work, so that wouldn't add much value - but hey, just what i won't do...

that's technically trunk, but hey. note, i'm not saying you all are doing something wrong - it might be something with linux versions etc.

do you have selinux/grsecurity or like installed and active, btw ?

[jeenam]

richlv,

After doing a little more research it seems the rpm's that I've been using from here are the source of the problem:

http://download.opensuse.org/repositories/home:/ericgearhart:/zabbix/CentOS_5/

The self compiled binary that I installed on the zabbix server has no problems retrieving vfs.dev.read.

Thank you for your feedback.

[richlv]

After doing a little more research it seems the rpm's that I've been using from here are the source of the problem

interesting. would be useful to find out why...

[freakyflow7]

Hey guys, the version of ZABBIX server I'm running was compiled on the system its running on, we're not using any distribution RPMs for the install but still face the same problem with the disk information. IS there anything aside from the agent required to get disk info?

[richlv]

so do you have selinux/apparmor/grsecurity or similar active ?

[freakyflow7]

Hey Rich,

Nope, we don't have that enabled at all. These systems are built via a kickstart server and that's something that has been turned off in the build scripts.

[jeenam]

We also have a few clients that list the incoming bytes per second of the eth0 interface as 'unsupported'. At this point I've narrowed it down to all of the clients that have this problem as being Centos 5.2 machines. The machines we have running Centos 5.4 are properly returning data.

What distro and version are you running? Perhaps try using an older agent such as the 1.6.x code, or even 1.8.0.

[freakyflow7]

CentOS 5.5 here.

[Tristan]

Same problem here:
i have compiled agent 1.6.8, 1.8.4, 1.8.9 and 1.8.4 and when i execute the following command i get a [m|Collector is not started!] message

./zabbix_agentd -t vfs.dev.read[sda]

The same command on a fresh compiled zabbix agent 1.4.6 works perfect:

Does anybody know whats wrong? I monitor 250 hosts, and all the 1.6.x agents don't return any data.