odd problem with distributed erlang system

[Jenny Dybedahl]

Garry Hodgson <garry@REDACTED> writes:

>    - why only the second and subsequent subservers have this
>    problem? after all, by that time, both the master node and one
>    sub node have been logging successfully, with the master owning
>    the file system that is nfs-mounted on the subnodes, and the
>    first subnode accessing it via nfs.
> 
>    - why we have no other apparent nfs issues? the subnodes mount
>    /usr/local, which is where all the code, libs, scripts, and even
>    erlang itself lives.
> 
>    - why we don't see this on any other installation. it smells of a
>    configuration problem, yet this is the only symptom we see. no
>    errors show up in any of the system logs.
> 
> for what it's worth, i'm opening the logs using the following. it
> gets called by the master and the subservers when they start. am i
> doing something exceptionally stupid here?

Are all nodes using the same logfile, or are they using separate
logfiles stored on the same disk? If they're trying to use the same
log file, I'd say you have a classic case of NFS locking problem.

File locking over NFS is notoriously problematic. If there's any way
you can avoid having to write to the same file from two different
servers, do so.

-- 
"I live in the heart of the machine. We are one."