Re: ceph.conf

Xiubo Li <xiubli@xxxxxxxxxx> · Thu, 25 Aug 2022 10:10:23 +0800



    On 8/24/22 9:16 PM,
      Loreth.Andreas@xxxxxxxxxxxxxx wrote:

    
        Dear Madam and Sir,
         
        We have an HPC based on
            Centos 7 and a Ceph Storage Cluster in version 15.2.16
            octopus (stable).
        The Ceph Storage Cluster
            provides us with CephFS.
        We had mounted CephFS
            using the Kernel Driver with the noacl attribute, but then
            realised that we needed ACL for rootless Docker.
        By just removing the
            noacl attribute, CephFS was mounted with ACL and rootless
            Docker works as expected.
        On the cluster, this
            operation is not handled manually but instead by AutoFS.
        Our service provider has
            added the corresponding entry for AutoFS (noacl removed) but
            also added the following entries to ceph.conf.
         
        [client]
        fuse_default_permission=0
        client_acl_type=posix_acl
         
      
    This is only for user space cephfs clients. For kernel client you
      just need to remove the 'noacl' option in the mount command.

    
        I was able to mount
            CephFS with ACL without changing the ceph.conf and when
            adding noacl to the mount call again, CephFS is mounted
            without ACL despite the entries in ceph.conf.
        So what exactly is the
            function of these entries?
      
    
    The kernel client won't depend on ceph.conf to get the acl
      option. You must specify most of the options in the mount command:
    ... -o noacl,wsync ...
    There have two exceptions: mon_addr and secret options. If these
      two options are absent in the mount command it will try to get
      them from the ceph.conf.

    
    More detail please see
      https://docs.ceph.com/en/quincy/cephfs/mount-using-kernel-driver/.

    
        And one more follow-up
            question: When and by whom is ceph.conf read in on a client
            and can the re-reading be triggered manually on the client?
         
      
    The ceph.conf file will be loaded and parsed when most of the
      user space ceph daemons are starting. Such as the mds, osd, mon,
      ceph-fuse fs client daemons, etc.
    For the cephfs kernel client, when mounting the 'mount.ceph' will
      also load and parse it to get some info from it, such as the 'mon_addr' and 'secret'
        options as mentioned above.
    You cannot reload
        it as I know once the daemon is started, if you have changed it
        you must reboot the daemon. Instead of restarting the daemon,
        some options allow to be changed in runtime by using 'tell'
        command, more detail please see ceph docs.

      
    Thanks!

    
    -- Xiubo

    
        I would be pleased to
            receive information that clarifies these questions.
         
        Best regards,
         
        ^{Andreas
              Loreth}
        --
        Andreas
              Loreth
        Institut
              für Biophysikalische Chemie
              
        Medizinische
              Hochschule Hannover (MHH)
        OE
            4350, Carl-Neuberg-Str. 1, 30625 Hannover,
            Deutschland
            
        Tel.:  +49 511 532- 8718  Fax.: +49 511 532-
            5966
        loreth.andreas@xxxxxxxxxxxxxx
          
        www.mhh.de/bpc 
            
         
      _______________________________________________
Dev mailing list -- dev@xxxxxxx
To unsubscribe send an email to dev-leave@xxxxxxx

    
_______________________________________________
Dev mailing list -- dev@xxxxxxx
To unsubscribe send an email to dev-leave@xxxxxxx