SLES9 SP2 and ZBX_NOTSUPPORTED

[ReefShark]

I have a problem that is driving me nuts. I am installing and configuring Zabbix to monitor a mixture of linux, snmp and windows boxes. I currently run version 1.1beta4 on the server, and on the linux agents and things are running smoothly. Except for the agent running on the server. On december 7th, it had suddenly descided that the agent on the server shouldn't be able to retrieve certain information, like free disk space, file checksums, netload, etc.

After trying to figure out why this particular server had this problem (and only after 7 december 2005) I finally took a peak on a simular server, running the same SLES version, service pack and patchlevel. And indeed, same problem there.

So to make a long story short: SLES9 SP2 current patchlevel breaks most of the functionality of the Zabbix agent! Unfortunately, I am unable to determine what patch is responsible. I am however interested in any solution, because I really want to use Zabbix to monitor our hardware and 80% of our servers are SLES9 SP2 machines.

[Alexei]

Perhaps the service pack sets more strict access right to /proc filesystem. Can an ordinary user rfead files under /proc?

Please, also give me output of zabbix_agentd -p from SLES9 SP2.

[ReefShark]

I logged onto the system as the zabbix user. I went into proc and could cat files to my hearts content.

Output of zabbix_agentd -p

system.localtime [u|1135935434]
vfs.file.exists[/etc/passwd] [u|1]
vfs.file.time[/etc/passwd,modify] [u|1135341571]
vfs.file.size[/etc/passwd] [u|1231]
vfs.file.regexp[/etc/passwd,root] [s|root]
vfs.file.regmatch[/etc/passwd,root] [u|1]
cpu[idle1] [m|ZBX_NOTSUPPORTED]
io[disk_io] [d|20419408.000000]
kern[maxfiles] [u|314552]
memory[buffers] [u|149831680]
system[uname] [s|Linux stitches 2.6.5-7.193-bigs mp #1 SMP Wed Jul 20 14:39:18 UTC 2005 i686 i686 i386 GNU/Linux]
sensor[temp1] [m|ZBX_NOTSUPPORTED]
swap[total] [u|4299112448]
version[zabbix_agent] [s|1.1beta4]
agent.ping [u|1]
agent.version [s|1.1beta4]
kernel.maxfiles [u|314552]
kernel.maxproc [m|ZBX_NOTSUPPORTED]
vfs.file.cksum[/etc/services] [u|3307438535]
vfs.file.md5sum[/etc/services] [s|2e1d021c991a25ce7c834551ca6c6cf 8]
system.cpu.switches [m|ZBX_NOTSUPPORTED]
system.cpu.intr [u|2803812738]
net.tcp.dns[127.0.0.1,localhost] [u|0]
net.tcp.listen[80] [m|ZBX_NOTSUPPORTED]
net.tcp.port[,80] [u|1]
net.tcp.service[ssh,127.0.0.1,22] [u|1]
net.tcp.service.perf[ssh,127.0.0.1,22] [d|0.001040]
net.if.in[lo,bytes] [u|210827769]
net.if.out[lo,bytes] [u|210827769]
net.if.collisions[lo] [u|0]
vfs.fs.size[/,free] [u|1802572]
vfs.fs.inode[/,free] [u|0]
vfs.dev.read[sda,operations] [u|667262]
vfs.dev.write[sda,sectors] [u|194207036]
vm.memory.size[free] [u|83169280]
proc.num[inetd,,] [u|0]
proc.mem[inetd,,] [u|0]
system.cpu.util[all,user,avg1] [u|41]
system.cpu.load[all,avg1] [d|2.310000]
system.swap.size[all,free] [u|4299112448]
system.swap.in[all] [m|ZBX_NOTSUPPORTED]
system.swap.out[all,count] [m|ZBX_NOTSUPPORTED]
system.hostname [s|stitches]
system.uname [s|Linux stitches 2.6.5-7.193-bigs mp #1 SMP Wed Jul 20 14:39:18 UTC 2005 i686 i686 i386 GNU/Linux]
system.uptime [u|1988077]
system.users.num [d|1.000000]
mysql.version [s|mysql Ver 12.22 Distrib 4.0.18 , for suse-linux (i686)]

I also noticed some of the items do seem to return data to the server, but it is n't processed or stored correctly (noticed by checking Overview -> Latest data). For instance the MySQL Version is a blank field in Latest Data. I'm using the default templates.

Next week I am installing a new server (same type hardware) and I will try to determine which patch is breaking the agent. As soon as I know more, I will post it here.

[Alexei]

Everything looks fine. Perhaps it is because of new naming convention for items? Note that starting from 1.1beta3 item keys are called differently, however backward compatibility remains.

[ReefShark]

Ironically, I don't have the problem with several other clients running on SLES9 SP 2, but not the patchlevel of the two 'broken' clients.

I'll try to reinstall the entire thing when 1.1b5 is released, that will clear up any experiments and such I've been doing while exploring the app in the database and give me a clean and fresh start.

[mconigliaro]

i can confirm the issue with network load items on 1.1beta5 with sles9 sp2, but its only happening on one of my boxes. for what its worth, its my only 64 bit machine.

[ReefShark]

I've noticed the agents on the specific machines I mentioned aren't as backwards compatible as they should be.
I did a complete reinstall of zabbix server (including a clean database) and I am designing my own templates with the new Item syntax (the default Zabbix templates are still old syntax). Seems to have solved the problem for me so far.
(we run 32bits SLES9 btw)