question about striped missing data on certain devices

[mithandra]

Hi All,

We currently have 2 servers one running cacti 8.7b and a new cacti server running 8.7g

I hasve migrated the rrd files across to the new system using the following commands.

Code: Select all

export graphs:

/usr/bin/rrdtool dump /home/<username>/scripts/data/rrdfilename > /home/<username>/scripts/data/xml/rrdfilename.xml

import Graphs:

/usr/bin/rrdtool restore --force-overwrite /home/<username>/scripts/data/xml/rrdfilename.xml

All seem to we working perfectly appart from we have 2 devices that have stripped missing data.

Anyone got any suggestions as to what this could be

We have tried the cmd polller and have updated this to spine to see if it was a poller problem but its still aparent with spine also

On the old Cacti system which is also polling the devices these are working perfectly also they are polling using 32bit were as the new system I have built I am using 64 bit counters for all FastEthernet and Gig interfaces.

Thanks Mithandra

[G1]

Check for any duplicate data sources still active in background....that is the most probable reason.

[mithandra]

the system has been built from scratch all automatically from a database which holds snmp info of each device.

its all automatically generated so I should imagine that there would be no duoplicate data sources.

Could this be a problem from the migration?

thanks Mithandra

[TheWitness]

Compare the snmp timeout values between the two systems, compare the rrdtool ds max values on both systems, compare steps and heartbeats for both systems. If would have been easier taking a dump from the old system and migrating the database, followed by the xport/import rrdtool operation IMHO.

I would start over....

TheWitness

[mithandra]

Hi Thanks for the reply.

I have checked pretty much all settings against the old system and they are exactly the same - the snmp timout is set to 2000 on both machines

The Max VAlue for Data source item field in and out are set to 10000000

which is the default i think.

I was given the task to migrate the customer from one box to the latest version of cacti.

this has been done but also we are now automatically provisioning graphs from the custom IPS system which I have written.

The problem with the old cacti system is that most of the data was not verified and there was a lot of crap in there and we were not sure of the bandwidth statements.

All bandwidth statements in the new system are automatically detected from the MIB / CAR MIB if rate limited, so from end to end there can be no human error, as long as they have filled out the port description on the device then the interface will be monitored with the right graph title and the right interface speeds.

I have attached 2 graphs which are for the exact same interface on the same device but running from the old and new cacti systems.

[mithandra]

bump - anyone?

[TheWitness]

They both look good to me.

[mithandra]

Hi there,

Not sure what your refering to mate but you can see on the top graph that there is a missing data strip adn the bottom graph does not show this.

We are seeing loads more graphs also which has missing data. I have just learned that this machine is running on a VMWare machine, could this be the casue of the missing data?




thanks Mithandra

[TheWitness]

Are you talking about the red (what I assume is an hrule) on the new system? HRules are a bit of a pain. When RRDtool does it's calculation of the auto-scale, it does not take into account any HRULES, that's why the 'stripes' as you call them (sorry, it's a minor variation to the english language) is missing.

Just like a boot is not something you wear, but a place where you put your smelly shoes when you take a long road trip

TheWitness