Additional info: The feature documentation would include Install and config, active standby config, FUSE client, Kernal clients
the following has to be documented: 1. As per this defect :https://bugzilla.redhat.com/show_bug.cgi?id=1342751. setfacl has to be enabled on ceph-fuse client and ACL is not supported for kernel client as of now. 2. In ceph-fs by default, snapshots are not enabled. this is has to be enabled using the command :"ceph mds set allow_new_snaps true --yes-i-really-mean-it" 3. creation of multiple ceph-fs is diabaled. check the output when tried to create a new ceph-fs as below : >>ceph fs new myceph cephfs_metadata cephfs_data >>Error EINVAL: Creation of multiple filesystems is disabled. To enable this experimental feature, use 'ceph fs flag set enable_multiple true --yes-i-really-mean-it'
Hi, We (the CephFS development team) have only just become aware of this documentation's existence. Here are some important pieces of setup that are missing and need to be added: * How to create proper auth caps for clients (i.e. *not* telling people to copy their admin key to client nodes) * How to specify multiple mon IPs in mount paths (this should be the default in all examples, as passing only one is a weird/rare situation) * Client (fuse) config options like the ACL ones Rakesh mentions, and also things like setting "client quota = true" to enable quotas. Rakesh mentions the snapshot/multi-fs/multi-mds experimental feature flags above. This stuff needs to go into a "Limitations" section that explains that these features are disabled by default for a reason (do not simply tell the user to set the flags in order to use the feature without also explaining that the feature should not be used in general). Is the asciidoc source for this in e.g. github somewhere that we can comment on it directly? There are various mistakes in the text that I would like to annotate using something more convenient than typing into bugzilla. Once the setup/config aspects are done, the next highest priority for me would be adding troubleshooting guidance for things like crashed MDSs/clients, stuck/unresponsive clients, and the various health messages that can appear in "ceph status" from MDSs.
Oh, it looks like there is a lot of content there that isn't built into the single HTML page that I just reviewed from the link above
HI Bara, The scope for ceph-fs QE tech preview is installation and configuration along with : 1. ceph-mds active/passive standby 2. ceph-clients : fuse and kernel 3. repair tools And is it possible to move the bug status from On_QA to Assigned, we may see issues in future,( but as of now i am not sure).
doc changes : 1. need to add information of enabling ceph-tools 2. need to add information on installing ceph-fuse : yum install ceph-fuse. [after enabling ceph-tools] 3. section 4.1: Path restriction: 1. the doc says to generate a new keyring and while mouting as shown in example ,we need to add keyring path or else it will search for ceph.client.<userid>.keyring . so we need to add --keyring=<path of keyring> example : ceph-fuse -n client.1 --keyring='<path of keyring>' /mnt/cephfs -r /home/cephf 4. mount kernel client: the syntax appears to be wrong, there is root path missing refere the upstream docs : http://docs.ceph.com/docs/master/cephfs/kernel/. sudo mount -t ceph 192.168.0.1:6789:/ /mnt/mycephfs -o name=admin,secret=AQATSKdNGBnwLhAAnNDKnH65FmVKpXZJVasUeQ== right now after -t mon address:ip there is no root path, this causes error. 5. mount fuse client: It is mentioned here again to generate keyring with the permissions, so when mounting, we need to specify with keyring or the keyring file should be in /etc/ceph/ceph.client.<user id>.keyring typo change : cluster.conf -> ceph.conf 6. /etc/fstab need to add "," in kernel client example after "_netdev"