[Gluster-devel] Re: Unexpected behaviour when self-healing

[Daniel Wirtz]

Hello Einar,


> Be careful to have the right time. Use ntpdate in the startup of both
> servers and clients, then have ntp running as server on all machines.
> Use one local server as ntp server for your cluster, getting his time
> from an official ntp server. Then adjust /etc/ntp.conf and put the local
> server as the only one at all your hosts. This will assure all machines
> are correct about time, wich they don't right now, according to the
> timestamp mentioned.
>
> Regards,
> Einar
>
All machines are synchronized correctly, I'm already using ntp for a long
time. There is nowhere else a 1970-01-01 etc. timestamp in the whole system,
but inside the underlying exported directories that the GlusterFS servers
use, some files have this timestamp - as described earlier. If this is not
intended explicitly, then I think this is a bug happening when files are
automatically replicated to other nodes.

regards
Daniel