Cron'ing Poller_Reindex_Hosts

[routenull0]

I am looking to cron the poller_reindex_hosts.php in order to keep our interface names updated on all the graphs. Our setup when building out a device in Cacti is to configure it for every port/interface to be graphed, whether up/down/used/unused. We are turning up/down interfaces pretty regularly and I know the best way to deal with this is to cron the reindex

Code: Select all

/usr/bin/php /var/www/cacti/cli/poller_reindex_hosts.php -id=All

Although I read that it wasn't a good idea to run the reindex if there was current polling going on. We are polling with spine on 1min intervals so I am looking on for some suggestions on how to best set this up without causing issues.

Reindexing every other day would be sufficient, but I was thinking nightly at fist.


Relevant Cacti Info:

Code: Select all

Date 	Wed, 21 Aug 2013 13:04:45 +0000
Cacti Version 	0.8.8a
Cacti OS 	unix
SNMP Version 	NET-SNMP version: 5.5
RRDTool Version 	RRDTool 1.3.x 

System Specs:

Code: Select all

CPU Info:
Intel(R) Xeon(R) CPU X5650 @ 2.67GHz stepping 02
(12 Total Cores)
Memory:
25GB

[gandalf]

Do the interface names change while the interface number remains constant?
In this case, you may define a different automatic reindex trigger to get things solved
R.

[routenull0]

Do the interface names change while the interface number remains constant?
In this case, you may define a different automatic reindex trigger to get things solved
R.

Correct, it is mostly interface name/descriptions updating from "DISABLED" or the link to a system hostname or circuit identifier.

[gandalf]

Bad luck. Renaming an active interface is not catched by our reindex methods, currently. We will improve this a bit, but your case is not catched by that move. The reason for this is, that a whole SNMP table scan is expensive related to only fetching a few data for those interfaces, that are graphed. So we try to avoid a full scan.
This is done only on a reindex trigger (e.g. when a host returns from a reboot, a new index is discovered or the like).

Currently, running the reindex script from the cli directory is the only method I'm aware of to be run in batch mode. Use "--help" to leran the options available.
R.

[routenull0]

Bad luck. Renaming an active interface is not catched by our reindex methods, currently. We will improve this a bit, but your case is not catched by that move. The reason for this is, that a whole SNMP table scan is expensive related to only fetching a few data for those interfaces, that are graphed. So we try to avoid a full scan.
This is done only on a reindex trigger (e.g. when a host returns from a reboot, a new index is discovered or the like).

Currently, running the reindex script from the cli directory is the only method I'm aware of to be run in batch mode. Use "--help" to leran the options available.
R.

That was my understanding as well, but the question I had was is there truly a problem with running the reindex script when there is a potential poller running as well?

[gandalf]

That was my understanding as well, but the question I had was is there truly a problem with running the reindex script when there is a potential poller running as well?

I'm sorry. Latest Cacti release is able to do this in parallel. It will increase the server load, of course.
R.

[routenull0]

That was my understanding as well, but the question I had was is there truly a problem with running the reindex script when there is a potential poller running as well?

I'm sorry. Latest Cacti release is able to do this in parallel. It will increase the server load, of course.
R.

0.8.8b allows it in parallel? Awesome!

[gandalf]

In parallel means: when the polling is running.
It does not mean: threaded (next move?)
R.

[tosage]

Sorry for the question Reinhard, but what do you mean by

"threaded (next move?)"

I have too thousands of equipments description to update and your discution interested me

David