Dear developers,
I find a bug in FileStore::mount() function. If we use raw partitial
as a Journal disk, like /dev/sda1, then it will cause a error when
setting the parameters "m_filestore_max_inline_xattr_size" and
"m_filestore_max_inline_xattrs".
First of all, when creating the FileStore, it calls the mount() and
detect fs_type as XFS, so it will check
m_filestore_replica_fadvise(true default) and set it to false. At the
same time, m_fs_tpye = FS_TYPE_XFS.
However, when create JournalStore, it also enters mount() function and
detect fs_type. Unfourtuanlly, this time, m_filestore_replica_fadvise
= false so that it cannot update m_fs_type = FS_TYPE_XFS. Instread, it
will sets to m_fs_tpye = FS_TYPE_OTHER. This bug is quite easy to
repair.
By the way, I wanna ask another question about XFS, I find that if I
use ceph cluster heavily for about 2-3 hour. XFS always wakes up a
kernel thread (xfsaild) to deal with its inodes reading and writing.
This will impact the performance a lot and make the cluster's
throughput drop heavily. So how to solve this problem and why this
occurs after running a heavily read, write operation after 2-3 hour.
Best Regards,
Neal Yao
--
To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
