Re: [PVE-User] Cephfs starting 2nd MDS
your .conf references mds.1 (id =1) but your command starts the mds with id=scvirt03 so the block in ceph.conf is not used. replace [mds.1] with [mds.scvirt03] btw: iirc you can not have just numerical id's for mds's for some versions now, so mds.1 would not be valid either. kind regards Ronny Aasen On 08. aug. 2018 07:54, Vadim Bulst wrote: Hi Alwin, thanks for your advise. But no success. Still same error. mds-section: [mds.1] host = scvirt03 keyring = /var/lib/ceph/mds/ceph-scvirt03/keyring Vadim On 07.08.2018 15:30, Alwin Antreich wrote: Hello Vadim, On Tue, Aug 7, 2018, 12:13 Vadim Bulst wrote: Dear list, I'm trying to bring up a second mds with no luck. This is what my ceph.conf looks like: [global] auth client required = cephx auth cluster required = cephx auth service required = cephx cluster network = 10.10.144.0/24 filestore xattr use omap = true fsid = 5349724e-fa96-4fd6-8e44-8da2a39253f7 keyring = /etc/pve/priv/$cluster.$name.keyring osd journal size = 5120 osd pool default min size = 1 public network = 172.18.144.0/24 mon allow pool delete = true [osd] keyring = /var/lib/ceph/osd/ceph-$id/keyring [mon.2] host = scvirt03 mon addr = 172.18.144.243:6789 [mon.0] host = scvirt01 mon addr = 172.18.144.241:6789 [mon.1] host = scvirt02 mon addr = 172.18.144.242:6789 [mds.0] host = scvirt02 [mds.1] host = scvirt03 I did the following to set up the service: apt install ceph-mds mkdir /var/lib/ceph/mds mkdir /var/lib/ceph/mds/ceph-$(hostname -s) chown -R ceph:ceph /var/lib/ceph/mds chmod -R 0750 /var/lib/ceph/mds ceph auth get-or-create mds.$(hostname -s) mon 'allow profile mds' mgr 'allow profile mds' osd 'allow rwx' mds 'allow' > /var/lib/ceph/mds/ceph-$(hostname -s)/keyring chmod -R 0600 /var/lib/ceph/mds/ceph-$(hostname -s)/keyring systemctl enable ceph-mds@$(hostname -s).service systemctl start ceph-mds@$(hostname -s).service The service will not start. I also did the same procedure with the first mds which is running with no problems. 1st mds: root@scvirt02:/home/urzadmin# systemctl status -l ceph-mds@$(hostname -s).service ● ceph-mds@scvirt02.service - Ceph metadata server daemon Loaded: loaded (/lib/systemd/system/ceph-mds@.service; enabled; vendor preset: enabled) Drop-In: /lib/systemd/system/ceph-mds@.service.d └─ceph-after-pve-cluster.conf Active: active (running) since Thu 2018-06-07 13:08:58 CEST; 2 months 0 days ago Main PID: 612704 (ceph-mds) CGroup: /system.slice/system-ceph\x2dmds.slice/ceph-mds@scvirt02.service └─612704 /usr/bin/ceph-mds -f --cluster ceph --id scvirt02 --setuser ceph --setgroup ceph Jul 29 06:25:01 scvirt02 ceph-mds[612704]: 2018-07-29 06:25:01.792601 7f6e4bae0700 -1 received signal: Hangup from PID: 3831071 task name: killall -q -1 ceph-mon ceph-mgr ceph-mds ceph-osd ceph-fuse radosgw UID: 0 Jul 30 06:25:02 scvirt02 ceph-mds[612704]: 2018-07-30 06:25:02.081591 7f6e4bae0700 -1 received signal: Hangup from PID: 184355 task name: killall -q -1 ceph-mon ceph-mgr ceph-mds ceph-osd ceph-fuse radosgw UID: 0 Jul 31 06:25:01 scvirt02 ceph-mds[612704]: 2018-07-31 06:25:01.448571 7f6e4bae0700 -1 received signal: Hangup from PID: 731440 task name: killall -q -1 ceph-mon ceph-mgr ceph-mds ceph-osd ceph-fuse radosgw UID: 0 Aug 01 06:25:01 scvirt02 ceph-mds[612704]: 2018-08-01 06:25:01.274541 7f6e4bae0700 -1 received signal: Hangup from PID: 1278492 task name: killall -q -1 ceph-mon ceph-mgr ceph-mds ceph-osd ceph-fuse radosgw UID: 0 Aug 02 06:25:02 scvirt02 ceph-mds[612704]: 2018-08-02 06:25:02.009054 7f6e4bae0700 -1 received signal: Hangup from PID: 1825500 task name: killall -q -1 ceph-mon ceph-mgr ceph-mds ceph-osd ceph-fuse radosgw UID: 0 Aug 03 06:25:02 scvirt02 ceph-mds[612704]: 2018-08-03 06:25:02.042845 7f6e4bae0700 -1 received signal: Hangup from PID: 2372815 task name: killall -q -1 ceph-mon ceph-mgr ceph-mds ceph-osd ceph-fuse radosgw UID: 0 Aug 04 06:25:01 scvirt02 ceph-mds[612704]: 2018-08-04 06:25:01.404619 7f6e4bae0700 -1 received signal: Hangup from PID: 2919837 task name: killall -q -1 ceph-mon ceph-mgr ceph-mds ceph-osd ceph-fuse radosgw UID: 0 Aug 05 06:25:01 scvirt02 ceph-mds[612704]: 2018-08-05 06:25:01.214749 7f6e4bae0700 -1 received signal: Hangup from PID: 3467000 task name: killall -q -1 ceph-mon ceph-mgr ceph-mds ceph-osd ceph-fuse radosgw UID: 0 Aug 06 06:25:01 scvirt02 ceph-mds[612704]: 2018-08-06 06:25:01.149512 7f6e4bae0700 -1 received signal: Hangup from PID: 4014197 task name: killall -q -1 ceph-mon ceph-mgr ceph-mds ceph-osd ceph-fuse radosgw UID: 0 Aug 07 06:25:01 scvirt02 ceph-mds[612704]: 2018-08-07 06:25:01.863104 7f6e4bae0700 -1 received signal: Ha
Re: [PVE-User] Cephfs starting 2nd MDS
Hi, On Wed, Aug 08, 2018 at 07:54:45AM +0200, Vadim Bulst wrote: > Hi Alwin, > > thanks for your advise. But no success. Still same error. > > mds-section: > > [mds.1] > host = scvirt03 > keyring = /var/lib/ceph/mds/ceph-scvirt03/keyring [mds] keyring = /var/lib/ceph/mds/ceph-$id/keyring So it will work for every MDS that you setup. Besides that, no extra options would be needed for MDS to start. More then the bellow lines should not be needed to get the mds started. mkdir -p /var/lib/ceph/mds/ceph-$SERVER chown -R ceph:ceph /var/lib/ceph/mds/ceph-$SERVER ceph --cluster ceph --name client.bootstrap-mds \ --keyring /var/lib/ceph/bootstrap-mds/ceph.keyring auth \ get-or-create mds.$SERVER osd 'allow rwx' mds 'allow' mon 'allow profile mds' \ -o /var/lib/ceph/mds/ceph-$SERVER/keyring If it's not working, whats the output of 'systemctl status ceph-mds@'? -- Cheers, Alwin ___ pve-user mailing list pve-user@pve.proxmox.com https://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-user
Re: [PVE-User] Cephfs starting 2nd MDS
Thanks guys - great help! All up and running :-) On 08.08.2018 09:22, Alwin Antreich wrote: Hi, On Wed, Aug 08, 2018 at 07:54:45AM +0200, Vadim Bulst wrote: Hi Alwin, thanks for your advise. But no success. Still same error. mds-section: [mds.1] host = scvirt03 keyring = /var/lib/ceph/mds/ceph-scvirt03/keyring [mds] keyring = /var/lib/ceph/mds/ceph-$id/keyring So it will work for every MDS that you setup. Besides that, no extra options would be needed for MDS to start. More then the bellow lines should not be needed to get the mds started. mkdir -p /var/lib/ceph/mds/ceph-$SERVER chown -R ceph:ceph /var/lib/ceph/mds/ceph-$SERVER ceph --cluster ceph --name client.bootstrap-mds \ --keyring /var/lib/ceph/bootstrap-mds/ceph.keyring auth \ get-or-create mds.$SERVER osd 'allow rwx' mds 'allow' mon 'allow profile mds' \ -o /var/lib/ceph/mds/ceph-$SERVER/keyring If it's not working, whats the output of 'systemctl status ceph-mds@'? -- Cheers, Alwin ___ pve-user mailing list pve-user@pve.proxmox.com https://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-user -- Vadim Bulst Universität Leipzig / URZ 04109 Leipzig, Augustusplatz 10 phone: ++49-341-97-33380 mail:vadim.bu...@uni-leipzig.de ___ pve-user mailing list pve-user@pve.proxmox.com https://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-user
Re: [PVE-User] How to use lvm on zfs ?
El 07/08/18 a las 17:51, Yannis Milios escribió: (zfs create -V 100G rpool/lvm) and make that a PV (pvcreate /dev/zvol/rpool/lvm) and make a VG (vgcreate pve /dev/zvol/rpool/lvm) and then a LV (lvcreate -L100% pve/data) Try the above as it was suggested to you ... But I suspect I have no space to create an additional zfs volume since the one mounted on "/" occupied all the space No, that's a wrong assumption, zfs does not pre-allocate the whole space of the pool, even if looks like it does so. In short there is no need to "shrink" the pool in order to create a zvol as it was suggested above... Still, the whole idea of having LVM ontop of ZFS/zvol is a mess, but if you insist, it's up to you ... A combination of Linux RAID + LVM would look much more elegant in your case, but for that you have to reinstall PVE by using the Debian iso. During the installation create a linux raid array with lvm on top and then add PVE repos ass described in the wiki: https://pve.proxmox.com/wiki/Install_Proxmox_VE_on_Debian_Jessie That's right. Now I understand that lvm/zfs would be a mess. Mainly because zfs doesn't create a block devices such as partitions on which I could do pvcreate ... and make It part of a lvm volumen group. After a (zfs create -V 100G rpool/lvm) a have to do a losetup to create a loop device an so on... Instead, I will keep zfs Raid mounted on "/" (local storage) on the last 4 Proxmox, remove the local-lvm storage from all Proxmox, and resize the local storage of the first 4 Proxmox . In such a way that all the 8 Proxmox have just local storage making the migration of VMs between nodes easy. ___ pve-user mailing list pve-user@pve.proxmox.com https://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-user ___ pve-user mailing list pve-user@pve.proxmox.com https://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-user
Re: [PVE-User] How to use lvm on zfs ?
Why does Proxmox team have not incorporated a software Raid in the install process ? So that we could include redundancy and lvm advantages when using local disks. El 08/08/18 a las 09:23, Denis Morejon escribió: El 07/08/18 a las 17:51, Yannis Milios escribió: (zfs create -V 100G rpool/lvm) and make that a PV (pvcreate /dev/zvol/rpool/lvm) and make a VG (vgcreate pve /dev/zvol/rpool/lvm) and then a LV (lvcreate -L100% pve/data) Try the above as it was suggested to you ... But I suspect I have no space to create an additional zfs volume since the one mounted on "/" occupied all the space No, that's a wrong assumption, zfs does not pre-allocate the whole space of the pool, even if looks like it does so. In short there is no need to "shrink" the pool in order to create a zvol as it was suggested above... Still, the whole idea of having LVM ontop of ZFS/zvol is a mess, but if you insist, it's up to you ... A combination of Linux RAID + LVM would look much more elegant in your case, but for that you have to reinstall PVE by using the Debian iso. During the installation create a linux raid array with lvm on top and then add PVE repos ass described in the wiki: https://pve.proxmox.com/wiki/Install_Proxmox_VE_on_Debian_Jessie That's right. Now I understand that lvm/zfs would be a mess. Mainly because zfs doesn't create a block devices such as partitions on which I could do pvcreate ... and make It part of a lvm volumen group. After a (zfs create -V 100G rpool/lvm) a have to do a losetup to create a loop device an so on... Instead, I will keep zfs Raid mounted on "/" (local storage) on the last 4 Proxmox, remove the local-lvm storage from all Proxmox, and resize the local storage of the first 4 Proxmox . In such a way that all the 8 Proxmox have just local storage making the migration of VMs between nodes easy. ___ pve-user mailing list pve-user@pve.proxmox.com https://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-user ___ pve-user mailing list pve-user@pve.proxmox.com https://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-user
Re: [PVE-User] How to use lvm on zfs ?
Am 08.08.2018 um 15:32 schrieb Denis Morejon: > Why does Proxmox team have not incorporated a software Raid in the > install process ? So that we could include redundancy and lvm > advantages when using local disks. Because ZFS offers redundancy and LVM features (and much more) in a more modern way, e.g. during a rebuild only used blocks need to be resilvered, resulting in much greater speed. ZFS is intended to entirely replace MD-RAID and LVM. Only drawback of ZFS is that it needs bare metal disk access and must not (or at least should not) be used with hardware RAID controllers. This makes it difficult to use with older hardware, e.g. HP ProLiants which only have HP SmartArray controllers as disk controllers. It is possible to put some RAID controllers in HBA mode, though ZFS docs advise against it. Andreas ___ pve-user mailing list pve-user@pve.proxmox.com https://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-user
Re: [PVE-User] How to use lvm on zfs ?
I'd say that it is more convinient to support one method. Also mentioned in this thread that zfs could be considered to be a successor of MDraid+LVM. It is still a debian system with cutom kernel and some pve packages on top, so You could do anything just like on any standard debian system. On 8/8/18 3:32 PM, Denis Morejon wrote: Why does Proxmox team have not incorporated a software Raid in the install process ? So that we could include redundancy and lvm advantages when using local disks. El 08/08/18 a las 09:23, Denis Morejon escribió: El 07/08/18 a las 17:51, Yannis Milios escribió: (zfs create -V 100G rpool/lvm) and make that a PV (pvcreate /dev/zvol/rpool/lvm) and make a VG (vgcreate pve /dev/zvol/rpool/lvm) and then a LV (lvcreate -L100% pve/data) Try the above as it was suggested to you ... But I suspect I have no space to create an additional zfs volume since the one mounted on "/" occupied all the space No, that's a wrong assumption, zfs does not pre-allocate the whole space of the pool, even if looks like it does so. In short there is no need to "shrink" the pool in order to create a zvol as it was suggested above... Still, the whole idea of having LVM ontop of ZFS/zvol is a mess, but if you insist, it's up to you ... A combination of Linux RAID + LVM would look much more elegant in your case, but for that you have to reinstall PVE by using the Debian iso. During the installation create a linux raid array with lvm on top and then add PVE repos ass described in the wiki: https://pve.proxmox.com/wiki/Install_Proxmox_VE_on_Debian_Jessie That's right. Now I understand that lvm/zfs would be a mess. Mainly because zfs doesn't create a block devices such as partitions on which I could do pvcreate ... and make It part of a lvm volumen group. After a (zfs create -V 100G rpool/lvm) a have to do a losetup to create a loop device an so on... Instead, I will keep zfs Raid mounted on "/" (local storage) on the last 4 Proxmox, remove the local-lvm storage from all Proxmox, and resize the local storage of the first 4 Proxmox . In such a way that all the 8 Proxmox have just local storage making the migration of VMs between nodes easy. ___ pve-user mailing list pve-user@pve.proxmox.com https://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-user ___ pve-user mailing list pve-user@pve.proxmox.com https://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-user ___ pve-user mailing list pve-user@pve.proxmox.com https://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-user
Re: [PVE-User] How to use lvm on zfs ?
> Why does Proxmox team have not incorporated a software Raid in the > install process ? Because we consider mdraid unreliable and dangerous. > So that we could include redundancy and lvm advantages > when using local disks. Sorry, but we have software raid included - ZFS provides that. ___ pve-user mailing list pve-user@pve.proxmox.com https://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-user