Support #14450
openceph-fuse does not work without loaded ceph/rbd kernel module writing big data
0%
Description
Hi,
i was installing different OS ( centos 6, centos 7, Debian 8 ).
I was installing ceph-fuse only ( hammer release, since hammer i use with the ceph cluster ).
I copied the keyring to /etc/ceph/ceph.client.admin.keyring
I configured the network with mtu 9000
I mounted fuse via:
#ceph-fuse -m 10.0.0.1:6789 /ceph-storage
Run:
dd if=/dev/zero bs=256M count=16 of=/ceph-storage/dd1
Result:
dd1 will have exactly 104726528 bytes size while the dd command never ends.
The ceph cluster might report:
2016-01-21 13:18:06.892010 7f71b4b4f700 0 log_channel(cluster) log [INF] : closing stale session client.21431412 10.0.0.91:0/8147 after 303.883060
Now i reboot.
I load
modprobe ceph
modprobe rbd
and mount again with ceph-fuse and started again dd.
This time:
dd if=/dev/zero bs=256M count=8 of=/ceph-storage/dd1
8+0 records in
8+0 records out
2147483648 bytes (2.1 GB) copied, 28.3892 s, 75.6 MB/s
Success.
This goes for any OS i tried.
Ceph-fuse successfully mount.
You can delete any size of data, you can write small data.
Writing bigger data will result after 104726528 in a stale.
If you load the rbd/ceph module AFTER the mount, same result -> stale.
If you load the rbd/ceph module BEFORE the mount, -> success
Is this a known issue ? Any chance to work around this ?
I need to use kernel 2.6.32 ( openvz ) and ceph was introduced 2.6.34 ( thank you merphy ! ) so i have to depend on ceph-fuse or go another way.
Updated by Greg Farnum about 8 years ago
- Tracker changed from Bug to Support
- Priority changed from High to Normal