KB450103 – CephFS

Last modified: October 23, 2019
You are here:
Estimated reading time: 2 min

Enable snapshots with

  • ceph mds set allow_new_snaps true –yes-i-really-mean-it

Manual Snapshots

To take a snapshot:

  • From a client where the cephFS is mounted, navigate to directory you want to snapshot. cd /mnt/cephfs/data
  • cd .snap
  • mkdir new_snap_03-02-18

To recover files from snapshot:

  • From a client where the cephFS is mounted, navigate to directory you want to recover the file.cd /mnt/cephfs/data
  • cd .snap
  • Find the correct snapshot, cd new_snap_03-02-18
  • Find the file(s) you want to recover
  • cp -R {file(s)} /mnt/cephfs/data

To delete snapshot:

  • cd /mnt/cephfs/data/.snap
  • rmdir new_snap_03-02-18

Automated Snapshots

Use cephfs-snap to automatically create and delete old snapshots.

Usage Example:

cephfs-snap /mnt/server hourly 24
would make a snapshot in /mnt/server/.snap/ called hourly_<date>_<time>
where <date> is on the form YYYY-MM-DD and <time> is on the form
HH:MM:SS. This format makes sure snapshot names sort correctly in
chronological order even when sorted alphabetically. The 24 newest
snapshots matching the prefix are kept around. The rest are deleted.

Designed to be used with /etc/cron.{hourly,daily,weekly,monthly}

Usage Example:

To take hourly snaps, and delete the old ones after 24, create a file "/etc/cron.hourly/cephfs-snap"
And place
cephfs-snap / hourly 24

To mount, specify which namespace if you have more than one filesystem, specify which one to mount using the mds_namespace option, e.g. -o mds_namespace=myfs.,, /ceph/ovirt   ceph    dirstat,name=admin,secretfile=/root/.ceph/admin.secret,noatime,_netdev,mds_namespace=cephfs   0 2,, /ceph/ssdfs   ceph    dirstat,name=admin,secretfile=/root/.ceph/admin.secret,noatime,_netdev,mds_namespace=ssdfs    0 2



Samba 4.8.X and onward has the vfs_ceph object included.

The most recent package in the CentOS repos is 4.6.

To use vfs_ceph to have SMBshare a ceph_fs volume without mounting in userspace first, you have to install samba 4.8.X from source.

Installation From Source

Get the most recent 4.8.X release from the Sambagit page.

wget https://github.com/samba-team/samba/releases

Install dependencies:

yum install python-devel libacl-devel openldap-devel pam-devel

Configure Samba for installation. Use “./configure –help” for a full list of options. By default, we exclude the packages needed for this install to function as a Domain Controller

./configure --without-ad-dc

Compile source


To install compiled source

make install

Export samba to PATH file

export PATH=/usr/local/samba/bin/:/usr/local/samba/sbin/:$PATH

Create systemd service files


If you get a message like the following:

 2018-03-09 15:12:20.458881 [INF]  Cluster is now healthy
 2018-03-09 15:12:20.458821 [INF]  Health check cleared: MDS_CLIENT_RECALL (was: 1 clients failing to respond to cache pressure)
 2018-03-09 15:12:20.445771 [INF]  MDS health message cleared (mds.0): Client cephsmb1.45lab.com failing to respond to cache pressure
 2018-03-09 15:11:56.495660 [WRN]  Health check failed: 1 clients failing to respond to cache pressure (MDS_CLIENT_RECALL)
 2018-03-09 15:11:56.445334 [WRN]  MDS health message (mds.0): Client cephsmb1.45lab.com failing to respond to cache pressure

First, check the kernel. The cephfs kernel driver in the 3.10 kernel is fairly out of date.
Use the el repo kernel-ml to update to the most recent Linux kernel.

If running updated kernel and messages persist, see the following:

Was this article helpful?
Dislike 0
Views: 971
Unboxing Racking Storage Drives Cable Setup Power UPS Sizing Remote Access