"Losing" historical values?

[dlindquist]

Hi,

I have two hosts set up for monitoring, using Template_Linux, and they both are reporting (quite nicely) CPU system/nice/user/idle times.

However, it appears that the box that the zabbix server is running on keeps "losing" history for these items -- graphing, "500 latest values", "all values in selected period - week", etc all show only the most recent set of values (I think perhaps since the last zabbix server restart?)

The other host I'm monitoring does *NOT* have this problem (it's running only zabbix_agentd).

Is there some reason why I should see this? I've checked the item for it's "keep history" option, and it's set to 90 days, as per default...

Any ideas?

TIA!

[dlindquist]

No-one has any ideas?

This is very frustrating -- the one host seems to keep only a few minutes of history for various items like CPU, and about 1/4 of 'disk used/free' items.

I see nothing in my logs, and nothing in examining the items (which are all stock, anyway).

TIA for any help!

[dlindquist]

Possible Solution:

I noticed that all (or apparently all) of the "items" that were giving me problems (and only those) had entries in the "housekeeping" table.

In the absence of any other advice or suggestions from more knowledgeable folks, I wiped out the entries from this table.

It *appears* (sitting at ~12 hours or so now) that the values that were consistently "losing" history are no longer doing so (as of about the time I wiped out this table).

What does this table do? Why should entries in this table cause the expiration of historical values so quickly?

Is the "90" days to keep history setting on the item being interpreted as "minutes" or "seconds" or something?

[Alexei]

The table contans IDs of deleted resources, for example, item IDs. Housekeeper periodically scans the table and removes associated data.

[dlindquist]

OK, that makes sense.

So any idea why:
a) Only a small percentage of 'items' from a single host would be present in this table, and
b) The presence of items in this table would cause premature erasure of historical values?

[dratliff]

I have had a very odd problem with values that were put into the history table. I have a trigger that updates the history table. I need to do this because I cannot force a timestamp with the data and the data comes in at various times from remote units. I could write a program to poll data, but I decided it is easier to just use a federated database engine to the table of data and put it in the zabbix database.

My problem is that I dumped all of my historical data, in proper format, into the history table. I can view the values of it, but it won't graph it. Any ideas? I changed my housekeeper values so that they keep the data indefinitely. Could the housekeeper table be the issue?

Thanks,

Dave