I believe I've recently spent some time with this issue, so I hope this is helpful. Apologies if it's an unrelated dm/udev/ceph-disk problem. https://lists.freedesktop.org/archives/systemd-devel/2017-July/039222.html The above email from last July explains the situation somewhat, with the outcome (as I understand it) being future versions of lvm/dm will have rules to create the necessary partuuid symlinks for dm devices. I'm unsure when that will make its way into various distribution lvm packages (I haven't checked up on this for a month or two actually). For now I've tested running with the new dm-disk.rules on the storage nodes that need it, which allowed ceph-disk to work as expected. Cheers Tom -----Original Message----- From: ceph-users [mailto:ceph-users-bounces@xxxxxxxxxxxxxx] On Behalf Of Fulvio Galeazzi Sent: 19 January 2018 15:46 To: ceph-users@xxxxxxxxxxxxxx Subject: Missing udev rule for FC disks (Re: mkjournal error creating journal ... : (13) Permission denied) Hallo, apologies for reviving an old thread, but I just wasted again one full day as I had forgotten about this issue... To recap, udev rules nowadays do not (at least in my case, I am using disks served via FiberChannel) create the links /dev/disk/by-partuuid that ceph-disk expects. I see the "culprit" is this line in (am on CentOS, but Ubuntu has the same issue): /usr/lib/udev/rules.d/60-persistent-storage.rules ..... # skip rules for inappropriate block devices KERNEL=="fd*|mtd*|nbd*|gnbd*|btibm*|dm-*|md*|zram*|mmcblk[0-9]*rpmb", GOTO="persistent_storage_end" ..... stating that multipath'ed devices (called dm-*) should be skipped. I can happily live with the file mentioned below, but was wondering: - is there any hope that newer kernels may handle multipath devices properly? - as an alternative, could it be possible to update ceph-disk such that symlinks for journal use some other /dev/disk/by-<whatever>? Thanks! Fulvio On 3/16/2017 5:59 AM, Gunwoo Gim wrote: > Thank you so much Peter. The 'udevadm trigger' after 'partprobe' > triggered the udev rules and I've found out that even before the udev > ruleset triggers the owner is already ceph:ceph. > > I've dug into ceph-disk a little more and found out that there is a > symbolic link of > /dev/disk/by-partuuid/120c536d-cb30-4cea-b607-dd347022a497 at > [/dev/mapper/vg--hdd1-lv--hdd1p1(the_filestore_osd)]/journal and the > source doesn't exist. though it exists in /dev/disk/by-parttypeuuid > which has been populated by > /lib/udev/rules.d/60-ceph-by-parttypeuuid.rules > > So I added this in /lib/udev/rules.d/60-ceph-by-parttypeuuid.rules: > # when ceph-disk prepares a filestore osd it makes a symbolic link by > disk/by-partuuid but LVM2 doesn't seem to populate /dev/disk/by-partuuid. > ENV{ID_PART_ENTRY_SCHEME}=="gpt", ENV{ID_PART_ENTRY_TYPE}=="?*", > ENV{ID_PART_ENTRY_UUID}=="?*", > SYMLINK+="disk/by-partuuid/$env{ID_PART_ENTRY_UUID}" > And finally got the osds all up and in. :D > > Yeah, It wasn't actually a permission problem, but the link just > wasn't existing. > > > ~ # ceph-disk -v activate /dev/mapper/vg--hdd1-lv--hdd1p1 ... > mount: Mounting /dev/mapper/vg--hdd1-lv--hdd1p1 on > /var/lib/ceph/tmp/mnt.ECAifr with options > noatime,largeio,inode64,swalloc > command_check_call: Running command: /bin/mount -t xfs -o > noatime,largeio,inode64,swalloc -- /dev/mapper/vg--hdd1-lv--hdd1p1 > /var/lib/ceph/tmp/mnt.ECAifr > mount: DIGGIN ls -al /var/lib/ceph/tmp/mnt.ECAifr > mount: DIGGIN total 36 > drwxr-xr-x 3 ceph ceph 174 Mar 14 11:51 . > drwxr-xr-x 6 ceph ceph 4096 Mar 16 11:30 .. > -rw-r--r-- 1 root root 202 Mar 16 11:19 activate.monmap > -rw-r--r-- 1 ceph ceph 37 Mar 14 11:45 ceph_fsid drwxr-xr-x 3 ceph > ceph 39 Mar 14 11:51 current > -rw-r--r-- 1 ceph ceph 37 Mar 14 11:45 fsid lrwxrwxrwx 1 ceph ceph > 58 Mar 14 11:45 journal -> > /dev/disk/by-partuuid/120c536d-cb30-4cea-b607-dd347022a497 > -rw-r--r-- 1 ceph ceph 37 Mar 14 11:45 journal_uuid > -rw-r--r-- 1 ceph ceph 21 Mar 14 11:45 magic > -rw-r--r-- 1 ceph ceph 4 Mar 14 11:51 store_version > -rw-r--r-- 1 ceph ceph 53 Mar 14 11:51 superblock > -rw-r--r-- 1 ceph ceph 2 Mar 14 11:51 whoami ... > ceph_disk.main.Error: Error: ['ceph-osd', '--cluster', 'ceph', > '--mkfs', '--mkkey', '-i', u'0', '--monmap', > '/var/lib/ceph/tmp/mnt.ECAifr/activate.monmap', '- -osd-data', > '/var/lib/ceph/tmp/mnt.ECAifr', '--osd-journal', > '/var/lib/ceph/tmp/mnt.ECAifr/journal', '--osd-uuid', > u'377c336b-278d-4caf-b2f5-592ac72cd9b6', '- -keyring', > '/var/lib/ceph/tmp/mnt.ECAifr/keyring', '--setuser', 'ceph', > '--setgroup', 'ceph'] failed : 2017-03-16 11:30:05.238725 7f918fbc0a40 > -1 filestore(/v > ar/lib/ceph/tmp/mnt.ECAifr) mkjournal error creating journal on > /var/lib/ceph/tmp/mnt.ECAifr/journal: (13) Permission denied > 2017-03-16 11:30:05.238756 7f918fbc0a40 -1 OSD::mkfs: > ObjectStore::mkfs failed with error -13 > 2017-03-16 11:30:05.238833 7f918fbc0a40 -1 ** ERROR: error creating > empty object store in /var/lib/ceph/tmp/mnt.ECAifr: (13) Permission > denied > > > ~ # blkid /dev/mapper/vg--*lv-*p* | grep > '120c536d-cb30-4cea-b607-dd347022a497' > /dev/mapper/vg--ssd1-lv--ssd1p1: PARTLABEL="ceph journal" > PARTUUID="120c536d-cb30-4cea-b607-dd347022a497" > ~ # ls -al /dev/disk/by-id | grep dm-22 lrwxrwxrwx 1 root root 11 > Mar 16 11:37 dm-name-vg--ssd1-lv--ssd1p1 -> > ../../dm-22 > lrwxrwxrwx 1 root root 11 Mar 16 11:37 > dm-uuid-part1-LVM-n1SH1FvtfjgxJOMWN9aHurFvn2BpIsLZi89GWxA68hLmUQV6l5oy > iEOPsFciRbKg > -> ../../dm-22 > ~ # ls -al /dev/disk/by-parttypeuuid | grep dm-22 lrwxrwxrwx 1 root > root 11 Mar 16 11:37 > 45b0969e-9b03-4f30-b4c6-b4b80ceff106.120c536d-cb30-4cea-b607-dd347022a > 497 -> > ../../dm-22 > ~ # ls -al /dev/disk/by-uuid | grep dm-22 ~ # ls -al > /dev/disk/by-partuuid/ | grep dm-22 ~ # ls -al /dev/disk/by-path | > grep dm-22 > > > Best Regards, > Nicholas Gim. > > On Wed, Mar 15, 2017 at 6:46 PM Peter Maloney > <peter.maloney@xxxxxxxxxxxxxxxxxxxx > <mailto:peter.maloney@xxxxxxxxxxxxxxxxxxxx>> wrote: > > On 03/15/17 08:43, Gunwoo Gim wrote: >> After a reboot, all the partitions of LVM don't show up in >> /dev/mapper -nor in the /dev/dm-<dm-num> or /proc/partitions- >> though the whole disks show up; I have to make the hosts run one >> 'partprobe' every time they boot so as to have the partitions all >> show up. > Maybe you need this after partprobe: > > udevadm trigger > > >> >> I've found out that the udev rules have never triggered even when >> I removed the DEVTYPE checking part; checked with a udev >> line: RUN+="/bin/echo 'add /dev/$name' >> /root/log.txt" >> I've also tried chowning all the /dev/dm-<num> to ceph:disk in >> vain. Do I have to use the udev rules even if the /dev/dm-<num> s >> are already owned by ceph:ceph? >> > No, I think you just need them owned by ceph:ceph. Test that with > something like: > > sudo -u ceph hexdump -C /dev/dm-${number} | head > > (which reads, not writes...so not a full test, but close enough) > > And also make sure the files in /var/lib/ceph/{osd,mon,...} are > owned by ceph:ceph too. Maybe you have a mix of root and ceph, which > is easy to cause by running it as root when ceph owns some files. > > > And FYI, I don't like udev, and did not use ceph-deploy or > ceph-disk. I did it with a very simple init script instead: > > >> #!/bin/bash >> mkdir -p /var/run/ceph >> chown ceph:ceph /var/run/ceph >> chgrp -R ceph /var/log/ceph >> for d in /var/lib/ceph/osd/*/journal; do >> d=$(readlink -f "$d") >> chown ceph:ceph "$d" >> done > > This works on ubuntu 14.04 as is, a badly written init script, but I > think centos will not accept it without the lsb tags. > > A side effect of doing it this way is you have to manually run the > script again when replacing or adding disks, since it is not run > after hot swap like udev is. > > > >> Thank you very much for reading. >> >> Best Regards, >> Nicholas. >> >> On Wed, Mar 15, 2017 at 1:06 AM Gunwoo Gim <wind8702@xxxxxxxxx >> <mailto:wind8702@xxxxxxxxx>> wrote: >> >> Thank you very much, Peter. >> >> I'm sorry for not clarifying the version number; it's kraken >> and 11.2.0-1xenial. >> >> I guess the udev rules in this file are supposed to change >> them : /lib/udev/rules.d/95-ceph-osd.rules >> ...but the rules' filters don't seem to match the DEVTYPE >> part of the prepared partitions on the LVs I've got on the host. >> >> Would it have been the cause of trouble? I'd love to be >> informed of a good way to make it work with the logical >> volumes; should I fix the udev rule? >> >> ~ # cat /lib/udev/rules.d/95-ceph-osd.rules | head -n 19 >> # OSD_UUID >> ACTION=="add", SUBSYSTEM=="block", \ >> ENV{DEVTYPE}=="partition", \ >> ENV{ID_PART_ENTRY_TYPE}=="4fbd7e29-9d25-41b8-afd0-062c0ceff05d", \ >> OWNER:="ceph", GROUP:="ceph", MODE:="660", \ >> RUN+="/usr/sbin/ceph-disk --log-stdout -v trigger /dev/$name" >> ACTION=="change", SUBSYSTEM=="block", \ >> ENV{ID_PART_ENTRY_TYPE}=="4fbd7e29-9d25-41b8-afd0-062c0ceff05d", \ >> OWNER="ceph", GROUP="ceph", MODE="660" >> >> # JOURNAL_UUID >> ACTION=="add", SUBSYSTEM=="block", \ >> ENV{DEVTYPE}=="partition", \ >> ENV{ID_PART_ENTRY_TYPE}=="45b0969e-9b03-4f30-b4c6-b4b80ceff106", \ >> OWNER:="ceph", GROUP:="ceph", MODE:="660", \ >> RUN+="/usr/sbin/ceph-disk --log-stdout -v trigger /dev/$name" >> ACTION=="change", SUBSYSTEM=="block", \ >> ENV{ID_PART_ENTRY_TYPE}=="45b0969e-9b03-4f30-b4c6-b4b80ceff106", \ >> OWNER="ceph", GROUP="ceph", MODE="660" >> >> >> ~ # udevadm info /dev/mapper/vg--ssd1-lv--ssd1p1 | grep >> ID_PART_ENTRY_TYPE >> E: ID_PART_ENTRY_TYPE=45b0969e-9b03-4f30-b4c6-b4b80ceff106 >> >> ~ # udevadm info /dev/mapper/vg--ssd1-lv--ssd1p1 | grep DEVTYPE >> E: DEVTYPE=disk >> >> >> Best Regards, >> Nicholas. >> > > -- > You can find my PGP public key here: https://google.com/+DewrKim/about > <http://google.com/+DewrKim/about> > > > _______________________________________________ > ceph-users mailing list > ceph-users@xxxxxxxxxxxxxx > http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com > _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
 |