Metric data not being charted past detailed cutoff date

[dr7d7]

Hi,
We are running Hyperic 4.1.0, and have the following settings:

Run Database Maintenance Every: 1 hours
Delete Detailed Metric Data Older Than: 7 days
Reindex Metric Data Tables Nightly: Yes
Delete Alerts Older Than: 36 days
Delete Events and Logs Older Than: 36 days

When we go to view the Indicator charts for a particular machine, we see the following behavior:

1. For a 7-day (time range) chart, we see lots of data points -- a beautiful chart
2. For an 8-day chart, we only see one or two data points, and it's relatively useless to interpret..

My hypothesis is that the "Delete Detailed Metric Data Older Than" setting has something to do with this. I would have expected that the amount of deleting wouldn't really affect how the chart looks though... Can anyone help out with this?

I've attached 2 screenshots (the 7-day chart, and the 8-day chart)

Thanks!

[dr7d7]

One more potentially useful piece of information: These symptoms seem mostly to appear with data collected at a relatively frequent interval (e.g. every 2 minutes). Other plots (for example "Availability") seems to display well when I select 8 days...

Any ideas?

[dr7d7]

<bump>

[dr7d7]

Just to clarify that I know that detailed metric data should be compressed after 7 days. That's not the issue here... The issue is that the charts don't show any data for days less than 7 days (i.e. the uncompressed data) when the 8-day chart is displayed.

Perhaps I should be posting this in the "Hyperic Bugs" forum instead...

[dr7d7]

Hi,
We are still seeing this issue. Could it be that the open source version of Hyperic really deletes all data after 7 days? I can't believe no one else is seeing this. Any tips would be greatly appreciated.
Thanks!

[dr7d7]

I just attached two more screenshots (7-day-bytes.png, and 8-day-bytes.png). These (in contrast to the previous screenshots) show that sometimes extending the chart date range results in the "No Metric Data Available" message.

I also attached a screenshot of my settings.

[iugav]

i justed started using hyperic, so dont have your problems yet... Anyway it is so strange, than nobody give any comment on these issue thread.

[Boris59]

I didn't get any earlier part of this thread, but I just checked with my
test system (4.1.2 OpenSource on Ubuntu Server x64) and it shows data
which is one month old. So the answer is no, data does not get cut after
seven days.

However, I sometimes have the impression that HQ (at least 4.1.0
OpenSource, which we run at the company for evaluation) is not capable
of showing too large time frames of data (such as 8 days) but will show
data within the same date area just fine by reducing the range to, say,
24 hours. Couldn't nail this one down properly yet, though, so I didn't
complain yet 🙂

Cheers, Bjoern

dr7d7 schrieb:
> Hi,
> We are still seeing this issue. Could it be that the open source version of Hyperic really deletes all data after 7 days? I can't believe no one else is seeing this. Any tips would be greatly appreciated.
> Thanks!
>

[Boris59]

Just checked the company installation and it does even go back to August 2008, which is when I installed the server (3.2.4 back then, upgarded to 4.1.0 on April this year).

Cheers, Bjoern

[dr7d7]

Hi Bjoern,
Thanks for the response. I wanted to clarify that I do see data for some metrics going back in time, but not others.

Here's an example of what I'm seeing when I'm in the "METRIC DATA" tab for a machine with "Free Memory", "Swap Used" and "Availability" as my metrics:

CASE 1 (selected "last 7 days" range):
-- I see data points for every single time slice, for all three metrics.

CASE 2 (selected "last 8 days" range):
-- "Availability" metric: I see data points for every single time slice
-- "Free Memory": I only see 2 data points, approximately 4 days ago.
-- "Swap Used": I only see 2 data points, approximately 4 days ago.

CASE 3 (selected "last 30 days" range):
-- "Availability" metric: I see data points for every single time slice
-- "Free Memory": I only see 2 data points, approximately 4 days ago.
-- "Swap Used": I only see 2 data points, approximately 4 days ago.

CASE 4 (selected "last 90 days" range):
-- "Availability" metric: I see data points for every single time slice
-- "Free Memory": I see 2 data points, approximately 4 days ago, and a cluster of points around 60 days ago.
-- "Swap Used": I see 2 data points, approximately 4 days ago, and a cluster of points around 60 days ago.

I previously hadn't tried to go back as far as 120 days in my previous tests, but I found it very interesting that data points started reappearing when I went that far back. Then I realized that's when I applied the patch (because I was seeing extreme slowness) as described in this post:

http://communities.vmware.com/thread/353546?tstart=0

I also noticed that another customer was complaining about losing data (or at least not having it displayed). The problem sounds similar to what I'm seeing.

So given the fact that:
1. I'm missing data starting around the time of the patch
2. Not many other people seem to be having this problem.

I'm starting to conclude that it was due to the patch...

So I'm on a "patched" 4.1.0 (build #1019)

[Boris59]

Hi.

I can't help you alot with a patched system as I (luckily) haven't got
far enough with HQ to require a patch 🙂 But as the forum post you
mentioned indicates the patch should be included in the 4.1.1 and onward
versions, you might just want to update your installation to 4.1.2 (came
out a few days ago). Should be a good idea to get away from a patched
system in any case.

Another thing I'd do in this case is using the HQ-API to fetch the
metric data of the resource and time range in question. I assume if any
data is there within the database, the HQ-API will be able to load it. I
actually did this when I discovered the "large time ranges not shown
properly" problem mentioned in my previous message, and HQ-API got me
all the data at all time ranges (i fetched the data daywise for about
three months worth of data).

As we'll reinstall HQ at the company sometime within the next weeks I'll
have an eye on this issue on HQ 4.1.2 and (finally!) a faster server.

Cheers, Bjoern