There have been several similar reports on the mailing list about this
[1][2][3][4] that are always a result of skipping step 6 from the Luminous
upgrade guide [5]. The new (starting Luminous) 'profile rbd'-style caps are
designed to try to simplify caps going forward [6].

TL;DR: your Openstack CephX users need to have permission to blacklist dead
clients that failed to properly release the exclusive lock.

[1]
http://lists.ceph.com/pipermail/ceph-users-ceph.com/2017-November/022278.html
[2]
http://lists.ceph.com/pipermail/ceph-users-ceph.com/2017-November/022694.html
[3] http://lists.ceph.com/pipermail/ceph-users-ceph.com/2018-May/026496.html
[4] https://www.spinics.net/lists/ceph-users/msg45665.html
[5]
http://docs.ceph.com/docs/master/releases/luminous/#upgrade-from-jewel-or-kraken
[6]
http://docs.ceph.com/docs/luminous/rbd/rbd-openstack/#setup-ceph-client-authentication


On Fri, Jul 6, 2018 at 7:55 AM Gary Molenkamp <[email protected]> wrote:

> Good morning all,
>
> After losing all power to our DC last night due to a storm, nearly all
> of the volumes in our Pike cluster are unmountable.  Of the 30 VMs in
> use at the time, only one has been able to successfully mount and boot
> from its rootfs.   We are using Ceph as the backend storage to cinder
> and glance.  Any help or pointers to bring this back online would be
> appreciated.
>
>   What most of the volumes are seeing is
>
> [    2.622252] SGI XFS with ACLs, security attributes, no debug enabled
> [    2.629285] XFS (sda1): Mounting V5 Filesystem
> [    2.832223] sd 2:0:0:0: [sda] FAILED Result: hostbyte=DID_OK
> driverbyte=DRIVER_SENSE
> [    2.838412] sd 2:0:0:0: [sda] Sense Key : Aborted Command [current]
> [    2.842383] sd 2:0:0:0: [sda] Add. Sense: I/O process terminated
> [    2.846152] sd 2:0:0:0: [sda] CDB: Write(10) 2a 00 00 80 2c 19 00 04
> 00 00
> [    2.850146] blk_update_request: I/O error, dev sda, sector 8399897
>
> or
>
> [    2.590178] EXT4-fs (vda1): INFO: recovery required on readonly
> filesystem
> [    2.594319] EXT4-fs (vda1): write access will be enabled during recovery
> [    2.957742] print_req_error: I/O error, dev vda, sector 227328
> [    2.962468] Buffer I/O error on dev vda1, logical block 0, lost async
> page write
> [    2.967933] Buffer I/O error on dev vda1, logical block 1, lost async
> page write
> [    2.973076] print_req_error: I/O error, dev vda, sector 229384
>
> As a test for one of the less critical vms, I deleted the vm and mounted
> the volume on the one VM I managed to start.  The results were not
> promising:
>
>
> # dmesg |tail
> [    5.136862] type=1305 audit(1530847244.811:4): audit_pid=496 old=0
> auid=4294967295 ses=4294967295 subj=system_u:system_r:auditd_t:s0 res=1
> [    7.726331] nf_conntrack version 0.5.0 (65536 buckets, 262144 max)
> [29374.967315] scsi 2:0:0:1: Direct-Access     QEMU     QEMU HARDDISK
> 2.5+ PQ: 0 ANSI: 5
> [29374.988104] sd 2:0:0:1: [sdb] 83886080 512-byte logical blocks: (42.9
> GB/40.0 GiB)
> [29374.991126] sd 2:0:0:1: Attached scsi generic sg1 type 0
> [29374.995302] sd 2:0:0:1: [sdb] Write Protect is off
> [29374.997109] sd 2:0:0:1: [sdb] Mode Sense: 63 00 00 08
> [29374.997186] sd 2:0:0:1: [sdb] Write cache: enabled, read cache:
> enabled, doesn't support DPO or FUA
> [29375.005968]  sdb: sdb1
> [29375.007746] sd 2:0:0:1: [sdb] Attached SCSI disk
>
> # parted /dev/sdb
> GNU Parted 3.1
> Using /dev/sdb
> Welcome to GNU Parted! Type 'help' to view a list of commands.
> (parted) p
> Model: QEMU QEMU HARDDISK (scsi)
> Disk /dev/sdb: 42.9GB
> Sector size (logical/physical): 512B/512B
> Partition Table: msdos
> Disk Flags:
>
> Number  Start   End     Size    Type     File system  Flags
>   1      1049kB  42.9GB  42.9GB  primary  xfs          boot
>
> # mount -t xfs /dev/sdb temp
> mount: wrong fs type, bad option, bad superblock on /dev/sdb,
>         missing codepage or helper program, or other error
>
>         In some cases useful info is found in syslog - try
>         dmesg | tail or so.
>
> # xfs_repair /dev/sdb
> Phase 1 - find and verify superblock...
> bad primary superblock - bad magic number !!!
>
> attempting to find secondary superblock...
>
>
>
> Which eventually fails.   The ceph cluster looks healthy, I can export
> the volumes from rbd.  I can find no other errors in ceph of openstack
> indicating a fault in either system.
>
>      - Is this recoverable?
>
>      - What happened to all of these volumes and can this be prevented
> from occurring again?  Note that any shutdown vm at the time of the
> outage appears to be fine.
>
>
> Relevant versions:
>
>      Base OS:  all Centos 7.5
>
>      Ceph:  Luminous 12.2.5-0
>
>      Openstack:  Latest Pike releases in centos-release-openstack-pike-1-1
>
>          nova 16.1.4-1
>
>          cinder  11.1.1-1
>
>
>
> --
> Gary Molenkamp                  Computer Science/Science Technology
> Services
> Systems Administrator           University of Western Ontario
> [email protected]                 http://www.csd.uwo.ca
> (519) 661-2111 x86882           (519) 661-3566
>
> _______________________________________________
> ceph-users mailing list
> [email protected]
> http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
>


-- 
Jason
_______________________________________________
ceph-users mailing list
[email protected]
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com

Reply via email to