Graphs changing when USB disk plugged in

[adam.morse]

Hi all. I've searched for a while on this issue but haven't found a solution (hope I've searched the right places!).

I have a new install of Cacti 0.8.7e on Ubuntu 9.04 which is working fine except for one area. I have a couple of Windows boxes that I monitor that occasionally have USB disks plugged into them for backup purposes. The problem I have is that as soon as the disk is attached, the drive letters change and presumably the partition list as SNMP sees it changes. The result is that the new disk (drive letter R:) takes the place of the Virtual Memory graph and the Virtual Memory graph takes the place of the Physical Memory one, with data for Physical Memory being lost. As soon as the disk is unplugged all is back to normal but this results in huge spikes making the memory graphs unreadable.

If I do a manual reindex of Get Mounted Partitions after plugging in the disk, the correct values appear and the graphs carry on as normal but I have to reindex again when unplugging. I am trying to find a way of auto reindexing the query so I don't have to remember to log in to Cacti each time I plug the disks in. I noticed the reindex method of the data sources for the hosts were the default "Uptime Goes Backwards", when "Index Count Changed" sounded more like it would fit the bill, so I updated the MySQL table host_snmp_query changing reindex_method to 2 for the hosts/queries for Get Mounted Partitions. Unfortunately, this did not make any difference and as far as I can tell, no reindex is being performed.

Is there a way of forcing a reindex of Get Mounted Partitions on each polling cycle for these hosts, or perhaps a more elegant way of resolving this?

Thanks in advance,
Ads

[TheWitness]

Verify all Indexes, or Index Count Changed will work for this. However, I am not sure we do a count of the number of indexes on the Host MIB.

TheWitness

[adam.morse]

I've just tried Verify All Fields as well but no luck. Checking the contents of host_snmp_cache before and after plugging the disk in and letting the poller run doesn't show any changes in the index numbers so I'm guessing it doesn't do a count.

Looking at ss_host_disk.php only tells me that I don't know much about PHP but there is a function ss_host_disk_reindex() which, not surprisingly, gets called on an index or query. I'm assuming if I managed to change the script to do a reindex on a get, this would dramatically slow down the polling process.

Do you have any pointers as to how I might be able to implement index count checking for this data source?

[Hyperlord]

I had this problem once, too and changing the reindex-method helped. I needed to reacreate the rrds AFAIK ... chose "index count may change" those days.

brgds