Empty graph notification

[hid3nax]

Hello everyone.
I have a really big Cacti setup with over 1000 hosts.
It's difficult to control these hosts and graphs.

The problem is that sometimes some queries for some hosts respond bad (e.g. apache stats) or doesn't respond at all (e.g. mysql stats). Then in graphs I have white space.
Is there any way to make an automatic detection of such graphs, which have stopped graphing so the server admin would/could get an email or a notice and could investigate why the graph isn't being showed?

Thanks.

[gandalf]

This might be achieved using THOLD plugin. But you've been warned: expect MANY messages!

Better is, to verify why dropouts occur. E.g. host unresponsive, polling time exceeded, rrd max for a data source exceeded, ...
R.

[hid3nax]

I was thinking about Thold too.
But....
Load average: 0.00 - matches thold.
Logged in Users 0 - matches thold
CPU Usage 0 - matches thold
VArious Apache, memcached, Nginx, Varnish, Mysql graphs.... Many values are 0 by nature so will match thold.

Anyway. I did manual investigation and fixed all bad graphs manually. Thanks anyway.

[mooneyn]

Also look into Autom8. It may be that your problem is that with so many devices, indexes are not being refreshed over time (such as new partitions, CPU, interfaces, controllers, drives) as they get added to your network inventory.

Autom8 will fix this, by using rules to create new graphs as new rows appear in your indexes.

Another tip is to periodically run the poller in MEDIUM logging for a few cycles. Search the cacti log for "value: U" and investigate the bad datasources. Usually its an interface or service that no longer exists in the table.

[hid3nax]

Yes, I know that. I've added the graph rules for such things (CPU, Interfaces, etc) and I run a cli script to reindex once a week.

Thanks for the last tip! Great idea.