Icinga2: Windows Disk performance data broken in InfluxDB

Created on 23 Mar 2018  路  13Comments  路  Source: Icinga/icinga2

Since the upgrade to Icinga 2.8.2, I get errors when sending metrics to InfluxDB of disk-windows:
partial write: unable to parse 'disk-windows,hostname=xxxx,service=disk,metric=C:\ crit=3758096384,max=75161927680,value=38654705664,warn=7516192768 1521812376': invalid tag format

Seems like the following commit broke it:
https://github.com/Icinga/icinga2/commit/7f92016e383252013e13aa3f17b7a6dc6424d3a0#diff-351369f332d56df0b38ba50ed7d32681L305

This breaks all our disks graphs for Windows!

areinfluxdb bug

Most helpful comment

Is there an option to issue an other 2.8.x release to fix this soon? we are affected by this too

All 13 comments

@N-o-X @gunnarbeutner

There's no need to highlight devs, we get notifications for this project.

Can you elaborate on the error message, what exactly is causing the invalid tag name error?

InfluxDB has issues with with slashes at the end. See https://github.com/influxdata/influxdb/issues/8587
These were changed to an underscore in 2.8.1, but the commit changes this. Which causes issues now :)

I see. What a mess.

@lippserd we should evaluate that next week during the dev sprint.

Backslashes could cause problems in several sql and nosql backends.
Would a configurable translation table (ie in feature config file) be an option?

Is there an option to issue an other 2.8.x release to fix this soon? we are affected by this too

Currently there are no plans for 2.8.x, this is why the issue targets 2.9.

When is 2.9 coming out? Disk-Windows is actually one of our most frequently used data sources (and saved our ass big time once). We are already missing one week. Can we somehow make the change ourselves without breaking the next update?

I have uploaded the check_disk binaries here. Replace the one on your windows machine with the right one (architectural differences) and when you install the update it will overwrite the custom one.

Make backups, just to be sure.

Hi, @Crunsher

I expecting a different performance output, but it looks like the same. c:\ 1234 B (1%); ....

cu denny

I see, I'm sorry. I thought this was a fix in the check_plugins, but actually we fixed the escaping in influxdb. You can install the snapshot packages on the Host that has the influxdb feature enabled, but the check_disk binaries provided by me do _not_ fix this.

hi,

O.K, than I have to wait for 2.9.x.

Fixed in 2.8.3

Working for me again.

Thanks!

Was this page helpful?
0 / 5 - 0 ratings