All of a sudden...graphs not graphing....SOLVED

[_JPL_]

None of my graphs are working anymore. As of Wednesday, March 14 at precisely 6PM all my graphs stopped producing data. There is nothing at all weird in my log files at that time to suggest something wrong. Every field has 'nan' in it. I haven't changed anything on the server at all. When I run the poller manually I can see that data is being gathered but it just isn't making it into the RRAs I guess? Any ideas what I can look for next? As I said, the poller in debug mode seems like everything should be working ok, just no data on the graphs.

Oh, I forgot. And new graphs aren't being created at all. Could the rra folder have somehow lost it's permissions?

drwxr-xr-x 2 cactiuser cactiuser 110592 Mar 2 12:05 rra

Those are the right perms, no? I hadn't changed anything at all so...?

Cacti version 0.8.6j
cactid version 0.8.6f

[maw]

did you remove your crontab entry?

EDIT:

From the "how-to" guide:

More often than not, this problem's cause is permissions. For instance if you run poller.php manually as root for testing, any .rrd files that it creates will be owned by root. Now when poller.php runs from cron, it will not be able to update these .rrd files because they are owned by root. You can double check the owner of your .rrd files by running the following command in your Cacti directory:

shell> ls -al rra/

[_JPL_]

I hadn't done anything at all which is the weird thing. I checked the perms/ownerships and they are right.

My owner is 'cactiuser' and the rra file perms are -rw-r--r--. The whole subtree of my cacti install is owned by 'cactiuser'.

Running the poller manually shows me that data is actually being gathered but it is not making it into the rra.

Where else can I look for clues? It simply just stopped working last week.

I know it sounds like a simple permissions thing (no data on graphs, no new graphs being created) but I hadn't changed anything.

I'm seeing a lot of WARNING: Result from SNMP not valid and WARNING: Result from script not valid in the log file. When run manually I can see that data is actually being received though.

[melchandra]

Check the poller_output table in the db. If it has more than a few thousand rows that's your problem. Empty out that table and your updated should resume. I'm not sure why but some bug in cacti makes it so not every row is removed during the polling cycle and so over time it builds up so that table is so large the sql query excedes the mem limit and crashes.

What I've done is then look at it after 1 polling cycle and run

SELECT * FROM `poller_output`,`poller_item` WHERE poller_output.local_data_id=poller_item.local_data_id

This will show you what polling items are getting "stuck" in the poller_output table.

[_JPL_]

When I ran

select count(*) from poller_output;

it returned some 1.2 million.

lol. That's gotta be the problem. I've truncated it using

truncate table poller_output;

from the mysql cli.

I should see a difference shortly.

EDIT:
That indeed was the problem. Thank you.