collectd runs as root unnecessarily
Bug #734179 reported by
Andrew Yates
This bug affects 4 people
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
collectd (Debian) |
New
|
Unknown
|
|||
collectd (Ubuntu) |
Confirmed
|
Undecided
|
Unassigned |
Bug Description
Binary package hint: collectd
The collectd init script runs collectd as root and does not provide a way to change the user in /etc/default/
Ubuntu 10.10
collectd version: 4.10.1-2
Changed in collectd (Debian): | |
status: | Unknown → New |
To post a comment you must log in.
On 14 March 2011 00:39, Sebastian Harl <email address hidden> wrote:
> Source: collectd collectd. Many of collectd's
> Version: 4.10.1-2.1
> Severity: wishlist
>
> Hi,
>
> (With this E-mail, I'm also submitting this as wishlist bug report to
> the Debian BTS and I'm going to fix it in my next upload. So, Ubuntu
> should get that as well on the next sync after that.)
>
> On Sun, Mar 13, 2011 at 07:04:39AM -0000, Andrew Yates wrote:
> > The collectd init script runs collectd as root and does not provide a
> > way to change the user in /etc/default/
> > plugins do not require root permissions, so USER and GROUP options
> > should be included in the init script and default file. (It might
> > also be a good idea to run collectd as an unprivileged user by default.
> > A quick test suggests that of the default plugins only 'df' may require
> > root privileges.)
>
> Fully agreed.
>
+1 from me there.
> collectd. conf) is the way to go.
> Also, I was thinking about (re)introducing a possibility to start
> several instances of collectd with different configurations from one
> init script, i.e. having the init script start one instance for each
> configuration file found in some subdirectory. I'm not yet sure, how to
> handle that properly. I guess, adding an option to /e/d/collectd to
> specify that directory (empty by default) and then ignoring any other
> files (including /e/collectd/
>
> Any comments, suggestions, questions about that?
>
Yes! That would be great.
The issue I have with collectd is that it forces one host name for all the
entries it reports about.
We need to track shared resources (SAN partitions) which get migrated among
the various RedHat Cluster Suite nodes together with an assigned Virtual IP
and have to keep monitoring them with their own host identifier (so it's
obvious from watching the graphs that we are still tracking the "df", for
instance, of the same SAN partition, no matter which RHCS node it is
currently mounted on). The only way I found so far to achieve that is to
start a separate collectd instance per such SAN disk (I haven't implemented
this, though).
A possibly better way would be to allow an entry to override the host name
it is reported under, and conditionally report it based on, e.g., whether
the given directory is a root of a file system or not.
Thanks,
--Amos