esxcfg-info output file not being process

[Stephen]

Hi,
I set up the vmktree and it is working for the most part. But I have
added a few ESX servers recently that don't work. I can see the
esxcfg-info output files from those servers in the /var/www/vmktree-
data/sys directory so the ssh and scp part of the data collection is
working. But there are no host-list files and no rrd files. I tried
to follow the vmktree-esx3-collector script to see where it fails but
jsut couldn't figure it out. Any suggestion would be appreciated.
Stephen

[Stephen]

I found out the problem is with my ESX servers. The information
spitted out by esxcfg-info was cut short by some errors during the
storage section. I work around the problem by modifying the vmktree-
esx3-collector file.

Replace this:
@getVSI=`$sshcommand $sudo /usr/sbin/esxcfg-info -a 2>&1`;

With this:
@getVSI=`$sshcommand '$sudo /usr/sbin/esxcfg-info -w ; $sudo /usr/sbin/
esxcfg-info -r ; $sudo /usr/sbin/esxcfg-info -s ; $sudo /usr/sbin/
esxcfg-info -n ; $sudo /usr/sbin/esxcfg-info -y ; $sudo /usr/sbin/
esxcfg-info -o' 2>&1`;

You can identify this problem by looking at the output file /var/www/
vmktree-data/sys/<servername>.txt file. If you see the file contains
only part of the esxcfg-info output, try the esxcfg-info command
locally on the ESX server. If my case, the command output cuts short
by the error:

Error: SysinfoException: Node
(VSI_NODE_storage_scsi_adapter_channel_target_lun_partition) ; Status
(bad001f)= Not supported; Message= Unable to Get Instance List

In case someone runs into similar problem, hope this help.

Stephen

[Lars Troen]

Good you figured out a workaround, Stephen. I have never seen that error message before, but it would be better if you could figure out *why* this is happening. Such errors may occur if there's a problem with your setup and we don't know what other side issues it could potentially cause.

By running esxcfg-info many times, collection time will probably increase quite a bit (last graph of your esx host)

Lars

[Stephen]

Hi Lars,
My "fix" doesn't really quite work. The data gets all mixed up. I
finally called VMWare support. We figured out what the error was and
corrected it on the ESX server.

Thanks a lot for sharing your wonderful tool!

Stephen

[Lars Troen]

Stephen,
Very good that you figured this out!

May I ask what the solution was? Others may come across the same issue too...

[Stephen]

The VMWare tech told me that it's a bug in ESX 3.5. When changes are
made to the iSCSI devices, the metabase may not be updated correctly,
ends up with invalid entries. In our case, rescanning the iSCSI bus
fixes the problem:

esxcfg-swiscsi -s

Stephen

[Stephen]

If you have fiber attached storage, use:

esxcfg-rescan <hba>

where <hba> is the name of your fiber channel HBA, e.g. vmhba0.