Munin fails to generate graph, stat should be less than end

[Rowan Wookey]

[Impact]

* If rrd data is older than the graph period i.e. a week the graph fails to generate with error 'start should be less than end'.

[Test Case]

[Regression Potential]

* This is a rather small and obvious bugfix from upstream. The only issue potentially is if a user manually backported the fix to their local code. Presuming they took the code directly from upstream, then there is no problem, as the new code will be identical to their manual change.

---

If rrd data is older than the graph period i.e. a week the graph fails to generate with error 'start should be less than end' issue has been reported and resolved upstream.

Upstream issue http://munin-monitoring.org/ticket/1358 and upstream fix http://munin-monitoring.org/changeset/57243edbb3ce8a40a774826b37a1852530778721/munin

The fix is minor, and effects trusty (I've not tested other versions).

[Launchpad Janitor (janitor)]

Status changed to 'Confirmed' because the bug affects multiple users.

[Hans-Joachim Kliemeck (bj7u6139zdyf-info-jjcftv6wldnz)]

please backport this bug

[gollum53 (smid)]

Yes, please backport this. If I correct the file myself, will this break something?
Thanx
Roman

[Ralf Heiringhoff (frosty-geek)]

Hi,

This bug still applies, can you please include this in the next maintenance update. The patch has been available upstream for 3 years now.

Regards,
Ralf

[Nish Aravamudan (nacc)]

This appears to have been fixed upstream in 2.20, which is in 16.04+.

[Nish Aravamudan (nacc)]

@frosty-geek: Would you be able to provide a clear test case in the description? Something we can use to recreate the issue and verify the fix works?

[Ralf Heiringhoff (frosty-geek)]

@nacc: Well one of our customers had the issue that munin-graph was producing serveral gigabytes (~3GB) of error logs within a very short time-frame (hours).

the server & client was the latest 14.04 available munin/munin-node package and the server was unable to poll the munin-node for several weeks, i guess that triggered the messages.

I'm attaching a munin-graph.log snippet & the corresponding rrd file.

----------------cut-------------------
2017/03/22 09:03:32 [WARNING] Could not draw graph "/data_XXX/var_cache_munin_www/host.fqdn.domain/mongodb201.host.fqdn.domain/diskstats_utilization/vg00_var-day.png": /data_XXX/var_cache_munin_www/host.fqdn.domain/mongodb201.host.fqdn.domain/diskstats_utilization/vg00_var-day.png
2017/03/22 09:03:32 [RRD ERROR] Unable to graph /data_XXX/var_cache_munin_www/host.fqdn.domain/mongodb201.host.fqdn.domain/diskstats_utilization/vg00_var-week.png : start (1489449812) should be less than end (1489147500)
2017/03/22 09:03:32 [RRD ERROR] rrdtool 'graph' '/data_XXX/var_cache_munin_www/host.fqdn.domain/mongodb201.host.fqdn.domain/diskstats_utilization/vg00_var-week.png' \
        '--title' \
        'Disk utilization for /dev/vg00/var - by week' \
        '--start' \
        '-12000m' \
        '--base' \
        '1000' \
        '--lower-limit' \
        '0' \
        '--upper-limit' \
        '100' \
        '--rigid' \
        '--vertical-label' \
        '% busy' \
        '--slope-mode' \
        '--height' \
        '175' \
        '--width' \
        '400' \
        '--imgformat' \
        'PNG' \
        '--lazy' \
        '--units-exponent' \
        '0' \
        '--font' \
        'DEFAULT:0:DejaVuSans,DejaVu Sans,DejaVu LGC Sans,Bitstream Vera Sans' \
        '--font' \
        'LEGEND:7:DejaVuSansMono,DejaVu Sans Mono,DejaVu LGC Sans Mono,Bitstream Vera Sans Mono,monospace' \
        '--color' \
        'BACK#F0F0F0' \
        '--color' \
        'FRAME#F0F0F0' \
        '--color' \
        'CANVAS#FFFFFF' \
        '--color' \
        'FONT#666666' \
        '--color' \
        'AXIS#CFD6F8' \
        '--color' \
        'ARROW#CFD6F8' \
        '--border' \
        '0' \
        '-W' \
        'Munin 2.0.19-3' \
        'DEF:autil=/var/lib/munin/host.fqdn.domain/mongodb201.host.fqdn.domain-diskstats_utilization-vg00_var-util-g.rrd:42:MAX' \
        'DEF:iutil=/var/lib/munin/host.fqdn.domain/mongodb201.host.fqdn.domain-diskstats_utilization-vg00_var-util-g.rrd:42:MIN' \
        'DEF:gutil=/var/lib/munin/host.fqdn.domain/mongodb201.host.fqdn.domain-diskstats_utilization-vg00_var-util-g.rrd:42:AVERAGE' \
        'CDEF:cutil=gutil,POP,UNKN' \
        'COMMENT: ' \
        'COMMENT: Cur\:' \
        'COMMENT:Min\:' \
        'COMMENT:Avg\:' \
        'COMMENT:Max\: \j' \
        'LINE1:gutil#00CC00:Utilization ' \
        'GPRINT:cutil:LAST:%6.2lf' \
        'GPRINT:iutil:MIN:%6.2lf' \
        'GPRINT:gutil:AVERAGE:%6.2lf' \
        'GPRINT:autil:MAX:%6.2lf\j' \
        'COMMENT:Last update\: Fri Mar 10 13\:15\:11 2017\r' \
        '--end' \
        '1489147500'
2017/03/22 09:03:32 [WARNING] Could not draw graph "/data_XXX/var_cache_munin_www/host.fqdn.domain/mongodb201.host.fqdn.domain/di...

[Nish Aravamudan (nacc)]

On Mon, Mar 27, 2017 at 11:52 AM, Ralf Heiringhoff
<email address hidden> wrote:
> @nacc: Well one of our customers had the issue that munin-graph was
> producing serveral gigabytes (~3GB) of error logs within a very short
> time-frame (hours).

Unfortunately, for SRU purposes
(https://wiki.ubuntu.com/StableReleaseUpdates) we need a (ideally
small) testcase to reproduce the issue.

In this case, the SRU team may make an exception, but I doubt it --
and it probably won't get accepted for SRU without the testcase.

-Nish

[Robie Basak (racb)]

If there's a reason there can be no test case then we can consider that, but I see no suitable reason described here. Please provide a test case, or if that is not feasible please explain why it is not feasible. Until then, there will be no update to 14.04.

[Robie Basak (racb)]

I'm rejecting this from the upload queue as there has been no answer.

[Nish Aravamudan (nacc)]

Thanks Robie, I would suggest this testcase, but it's so contrived, I'm not sure it's going to be reasonable to test:

Run Munin; get into a situaation where the rrd data is older than the graph period in Munin (e.g., some communication error to the Munin node) and Munin will throw an error.

But this is probably too vague to be a 'testcase' and without clearer steps for reproduction, it's not obvious how to SRU this change.