I'm also seeing messages such as the foloowing on my GlusterFS clients:
[2009-05-12 18:21:57] C [dht-diskusage.c:197:dht_is_subvol_filled]
mirrors: disk space on subvolume 'mirror1-1' is getting full (100.00 %), consider
adding more nodes
[2009-05-12 18:21:57] W [dht-diskusage.c:231:dht_free_disk_available_subvol] mir
rors: No subvolume has enough free space to create
[2009-05-12 18:21:57] W [dht-diskusage.c:231:dht_free_disk_available_subvol] mir
rors: No subvolume has enough free space to create
[2009-05-12 18:21:57] W [dht-diskusage.c:231:dht_free_disk_available_subvol] mir
rors: No subvolume has enough free space to create
This apparently started happening right from the beginning, when the disks
were completely empty.
Note that this is the first time I've tried using cluster/ha; could it be
failing to pass filesystem size info on to the other translators? I'm just
using cluster/ha to talk to the same server node on a different interface
(allows me to have an intact replicate setup, even with a switch failure,
so I don't have to worry about split-brain in that situation).
Note that df of the GlusterFS filesystem looks just fine, so it seems odd
that anything would be confused about filesystem size. Maybe it's just a
false warning, although it would eventually fill my logging space...
Thanks,
Brent
