Hi JC, Matt
Good news:
Last friday I first upgraded to 4.3.28, but the spiky behavior immediately
showed up. So I think this is not Xymon-version specific.
Then I did as JC suggested, dis/en-able debug & en/dis-abling the cache.
Since there is an SSD involved on my xymon server the impact is minimal and
there is no production running.
This fixes both issues!
1) The devmon/xymon related thing and the gaps for the graphs
disappears as *soon
as I disabled the caching* (--no-cache). As you say, not something I want
for long, but now we can have a specific look (A) why and (B) where caching
is a problem. I think that is good news!
2) I expect the memory leak error solved, as the release notes said, but
that will only show up over time (weeks).
3) The enable debugging showed me another problem in the, self-modified,
netapp.pl-script. I reverted my change and now there are no more spurious
xstatvolume,____-rrd-files anymore filling up my diskspace.
This is an error I introduced myself and mailed in November 2016 on the
list. Sorry for this.
Very happy now and hoping we can tackle the cache problem so I can enable
the launching of the rrd-deamons.
Peter
2017-03-21 16:36 GMT+01:00 Japheth Cleaver <user-87556346d4af@xymon.invalid>:
On Fri, Mar 17, 2017 at 8:56 AM, Peter Welter <user-f55666bd0d1e@xymon.invalid>
wrote:
Hi JC,
I'm still experiencing some difficulties with Xymon version
(4.3.27-1.el6.terabithia) software, that is being deployed from
http://terabithia.org/rpms/xymon/el6/i686/.
There are two different types of problems:
1) Has to do with the integration of Xymon/Devmon.
Although Devmon gets valid SNMP-data, for each poll, the values in
the if_load.Ethernet3_1.rrd-file (for example) are showing gaps. The next
value is so much larger than the rest, so the total graph is going beserk
because of the spikes that are being shown.
...[snip]
<!-- 2017-03-15 15:10:00 CET / 1489587000 -->
<row><v>5.7197560484e+01</v><v>5.7540255376e+01</v></row>
<!-- 2017-03-15 15:15:00 CET / 1489587300 -->
<row><v>5.8052253788e+01</v><v>5.7062462121e+01</v></row>
<!-- 2017-03-15 15:20:00 CET / 1489587600 -->
<row><v>5.8039204545e+01</v><v>5.7738579545e+01</v></row>
<!-- 2017-03-15 15:25:00 CET / 1489587900 -->
<row><v>5.8352395833e+01</v><v>5.7912187500e+01</v></row>
<!-- 2017-03-15 15:30:00 CET / 1489588200 -->
<row><v>5.7961458333e+01</v><v>5.8807500000e+01</v></row>
<!-- 2017-03-15 15:35:00 CET / 1489588500 -->
<row><v>5.7040675403e+01</v><v>5.7108262769e+01</v></row>
<!-- 2017-03-15 15:40:00 CET / 1489588800 -->
<row><v>5.7984999119e+01</v><v>5.8214662436e+01</v></row>
<!-- 2017-03-15 15:45:00 CET / 1489589100 -->
<row><v>1.6832224569e+16</v><v>1.6832224569e+16</v></row>
<!-- 2017-03-15 15:50:00 CET / 1489589400 -->
<row><v>4.4656922344e+16</v><v>4.4656922343e+16</v></row>
<!-- 2017-03-15 15:55:00 CET / 1489589700 -->
<row><v>5.7648150173e+01</v><v>5.7687031165e+01</v></row>
<!-- 2017-03-15 16:00:00 CET / 1489590000 -->
<row><v>5.9068884188e+01</v><v>5.9453689406e+01</v></row>
<!-- 2017-03-15 16:05:00 CET / 1489590300 -->
<row><v>NaN</v><v>NaN</v></row>
<!-- 2017-03-15 16:10:00 CET / 1489590600 -->
<row><v>NaN</v><v>NaN</v></row>
<!-- 2017-03-15 16:15:00 CET / 1489590900 -->
<row><v>NaN</v><v>NaN</v></row>
<!-- 2017-03-15 16:20:00 CET / 1489591200 -->
<row><v>NaN</v><v>NaN</v></row>
<!-- 2017-03-15 16:25:00 CET / 1489591500 -->
<row><v>NaN</v><v>NaN</v></row>
<!-- 2017-03-15 16:30:00 CET / 1489591800 -->
<row><v>1.9398478192e+07</v><v>1.8707899982e+07</v></row>
<!-- 2017-03-15 16:35:00 CET / 1489592100 -->
<row><v>5.6938284153e+01</v><v>5.6770437158e+01</v></row>
<!-- 2017-03-15 16:40:00 CET / 1489592400 -->
<row><v>NaN</v><v>NaN</v></row>
<!-- 2017-03-15 16:45:00 CET / 1489592700 -->
<row><v>NaN</v><v>NaN</v></row>
<!-- 2017-03-15 16:50:00 CET / 1489593000 -->
<row><v>NaN</v><v>NaN</v></row>
<!-- 2017-03-15 16:55:00 CET / 1489593300 -->
<row><v>NaN</v><v>NaN</v></row>
<!-- 2017-03-15 17:00:00 CET / 1489593600 -->
<row><v>NaN</v><v>NaN</v></row>
<!-- 2017-03-15 17:05:00 CET / 1489593900 -->
<row><v>NaN</v><v>NaN</v></row>
<!-- 2017-03-15 17:10:00 CET / 1489594200 -->
<row><v>NaN</v><v>NaN</v></row>
<!-- 2017-03-15 17:15:00 CET / 1489594500 -->
<row><v>NaN</v><v>NaN</v></row>
<!-- 2017-03-15 17:20:00 CET / 1489594800 -->
<row><v>NaN</v><v>NaN</v></row>
<!-- 2017-03-15 17:25:00 CET / 1489595100 -->
<row><v>3.5775056887e+07</v><v>3.4501518955e+07</v></row>
<!-- 2017-03-15 17:30:00 CET / 1489595400 -->
<row><v>5.7219344262e+01</v><v>5.7417704918e+01</v></row>
<!-- 2017-03-15 17:35:00 CET / 1489595700 -->
<row><v>5.7166338798e+01</v><v>5.9383825137e+01</v></row>
<!-- 2017-03-15 17:40:00 CET / 1489596000 -->
<row><v>5.6769617486e+01</v><v>5.6981202186e+01</v></row>
<!-- 2017-03-15 17:45:00 CET / 1489596300 -->
<row><v>5.7549617486e+01</v><v>5.7382732240e+01</v></row>
...[snip]
This behaviour does NOT occur on my current Xymon server (version
4.2.3) running on SLES11 SP4.
First I thought that this has to do with vmware, but that is not the
case. VM or bare metal; the behaviour is the same.
I made sure to see that even the devmon module is not causing the
problems. The same devmon software works fine on SLES and RHEL. The
snmpwalk-command does get valid SNMP-data, when writing to a files. It just
seems that Xymon does not update the rrd-file correctly!?!?
Any suggestions how to proceed?
Assuming that the numeric values are correct for the time periods that are
coming in, my first thought would be that there's something unusual going
on with RRD cacheing. Are you seeing this issue with other trends graphs,
either for other tests on this host, other hosts using this test/data, or
any other graphs period?
If it's unique to this, then that speaks to a problem with this specific
data transmission. If not, there could be a larger issue with xymond_rrd
(I/O performance, for example). I'd start with enabling debug output and
examining the logs for when it's receiving data for this test. (Not sure if
this is being sent via 'data' or 'status' messages, but you'll want to make
sure you're enabling debug for the right copy of xymond_rrd.)
If nothing there, then you might try disabling the cache, which will force
xymond_rrd to write things out as received (but will also increase I/O load
a lot).
If neither of those fix it, there could actually be an issue with the data
coming in. At about that point I would set up a channel listener looking
specifically for the host.svc messages related to this source so I could
physically see the contents of each one coming in and look for any
anomalies.
HTH,
-jc