Cacti EZ showing inconsistent total storage space

[citenx]

I'm having graphing issues with Cacti v0.8.8a deployed with Cacti EZ, NET-SNMP v5.5, RRD Tool v1.3.x.

My issue is that anything with the 'Get Mounted Partitions' data query attached is showing inconsistent space. Take a 7G partiton: for most poller cycles, the graph will show the total space of the partition to be 7G, then, seemingly at random, the graph will show 6.5G for a single poller cycle, then backup to 7G for a few, then show 12G:


Normally, the graph shows the correct data, but both the Total and Used graph data experiences this anomaly. I have an older version of Cacti monitoring the same hosts and it's graphs are solid. I also have this same VM at another site and it runs fine. I should also mention that the Realtime plugin view of the SNMP data is correct.

Thus far I have:

• Deleted and recreated a device
• Rebuilt poller cache
• Changed a host from SNMP v2 to v1
• Changed the timeout of a device to 5 seconds
• Checked the debug log for any anomalies and can't find any

Nothing has helped. Any help is appreciated. Thanks in advance.

[citenx]

Any suggestions?

[cigamit]

It pulls the total from the SNMP cache, you are seeing it lower, then back up to the normal because it is showing you an average. So it seems that the snmp cache is either being cleared or corrupted.

I wonder if it has anything to do with this bug.

0.8.8b
bug: Fix issue where host activate may flush the entire poller item cache

[citenx]

I can happily report that an upgrade worked! Thanks very much for your help!

The upgrade did cause another issue: thumbnails aren't rendering for 'Get Mounted Partitions' data query only. Screenshot attached. Any thoughts?

[cigamit]

You probably overrode the scripts / resources folder. CactiEZ uses a custom modification of the original data query that also adds a 'percentage' data point. You may want to restore the scripts and resources folders from backup.

[citenx]

No dice. I replaced both folders with the ones from backup and the inconsistent reporting returned and still no thumbnails. I've reverted back to the 8.8b versions of the scripts/resources folder.

[cigamit]

Even the ones from 0.8.8b are going to be wrong. You may need to redo the poller cache after restoring from backup.

[citenx]

Sorry, I should have been more clear. I have two instances: a version in DEV that was deployed from CactiEZ media. This was cloned and deployed to a second instance in LIVE.

The LIVE version had the inconsistent polling issues. These issues were rectified after running a 'yum update -y' but that broke thumbnails for the graphs associated with data query 'Get Mounted Partitions'.

To fix LIVE, I zipped the /var/www/html dir from DEV (unchanged since deployment) for backup, calling it cacti8.8a, performed the 'yum update -y' on DEV, zipped those contents to cacti8.8.b, then transferred both archives to LIVE.

The polling issue was fixed, so to fix the thumbnail issue I overwrote the scripts/resources folder with the contents from the cacti8.8a archive, which broke polling again and did not rectify the thumbnail issue, so I've reverted to the cacti8.8b scripts/resources folders.

Polling is working fine currently, but the thumbnails are still not working.

Again, thanks for your help!

[citenx]

This has been fixed. If anyone is having these issues, mine were three fold:

• Update CactiEZ (via yum), which upgraded Cacti from 8.8.a to 8.8b
• Consult the Spine documentation to tune the poller for my environment so it executed in less than a minute here
• Overwrote my version of rrd.php with the one from this thread. That fixed my thumbnails not drawing correctly

Thanks to everyone who has helped during this process. It was a tough one.