[collectd] Different OS for server and client

[Wallace, Ann]

On 2/2/10 1:13 PM, "Florian Forster" <octo at verplant.org> wrote:

> Hi Ann,
> 
> On Tue, Feb 02, 2010 at 11:10:50AM -0800, Wallace, Ann wrote:
>> Is it possible this is a byte order/endian issue?
> 
> yes, that sounds a bit like an endian issue. We used to have an endian
> problem with double values which was fixed in 4.2.5 and 4.3.1 (and all
> later versions).
> 
> Are *all* values sent from Linux to Solaris NaN or just some of them?
> Especially, can you check if you see a pattern between counter values
> (CPU usage, network and disk I/O) and gauge values (memory usage, DF
> graphs)?

Not all of the values from Linux to Solaris are NaN. CPU, Disk i/o, network
all work, df, users, load, uptime do not work.

> 
> Did you cross-compile the daemon? I.e. did you compile it on a x86
> machine for a SPARC architecture or something similar? In this case the
> configure script cannot determine the endianness of the target platform
> automatically.

I did not.

> 
> Is the data sent by Solaris machines stored correctly on the Linux
> server? This is a combination I have several times and know to work.
> Generally, at least, there's always room for more bugs I'm afraid ;)

Oddly, when I send Solaris (sparc) to Linux I see the UDP traffic but the
files never get created, and the user running collectd does have permission
to write to the rrd directory.

> 
> Regards,
> -octo