graph for nonexistent data source

[pwajda]

Hi,
I got graph on my host that draws IOs for dm-3 LVM device. The problem is that the host doesn't have dm-3 device.

How I could debug this situation?

note that 'dm-3' is set in data-source->index-value
and snmpwalk return only dm-0, dm-1, dm-2 (from LVM)

regards

[gandalf]

I don't understand. In case no dm-3 is present, you won't be able to graph it
R.

[pwajda]

You're absolutely right. Like I said before 'dm-3' is set and no 'dm-3' device is in snmpwalk results.
I might delete this graph but I want to know where data for the graph come from thus I could avoid such situations in the future.

Be aware that 'dm-3' device could exist on that node some time ago. I've taken over cacti's administration recently and I've just faced that problem.

[pwajda]

I found the source of the problem.

In poller cache items I got:

XXX-sql-node1 - Disk I/O - dm-3 SNMP Version: 1, Community: public, OID: .1.3.6.1.4.1.2021.13.15.1.1.3.28
RRD: /usr/local/share/cacti/rra/XXX-sql-node1_diskionwritten_5164.rrd

in fact under the .1.3.6.1.4.1.2021.13.15.1.1.3.28 OID aren't dm-3 device stats but sdb stats,
cause OID .1.3.6.1.4.1.2021.13.15.1.1.2.28 returns 'sdb'.

However, the question is why cacti keeps saying that it's dm-3 and how to force cacti to change it automatically (or to stop cheating!)?

[pwajda]

http://docs.cacti.net/manual:087:6_refe ... ndex_hosts

this is what I'm looking for (?)

if Index Type is diskIODevice is reindexing needed?

[gandalf]

Correct. That's a re-index issue. You may want to change reindexing to "index count changed". In case one index is added and another one deleted at the same time, this will unfortunately not help. So you might want to trigger a manual reindex on a scheduled base. Latest Cacti 087i is able to perform a complete reindex without trouble (but there's a bad bug when performing a cli based reindex on a single host; do NOT try that!).
R.

[pwajda]

So the best idea is to perform reindex on ALL hosts on specified data source (in this case: SNMP - Disk I/O (diskIOTable) (ks535) qid=11)?

Code: Select all

php -q poller_reindex_hosts.php --id=ALL --qid=11

Now I'm asking just for curiosity. I won't reindex it in this situation, cause it's safer for me to delete/modify those few graphs. Anyway, I wanted to know how to behave if problem will expand

Thanks!

[gandalf]

ATM, this would indeed be a good move.
Thinking more abouut re-indexing is always a good idea.
To be honest, I do NOT like our defaults ...
R.