Search results

IMHO this would be easier in that situation than trying to play catch with changing CDN IPs.

Single thread performance like you have with one disk inside one VM will always be disappointing, especially with small block sizes. You can just throw faster hardware at the problem. BTW: Your rados bench uses 16 parallel threads and a block size of 4MB. This will show you nearly the maximum...

Have you added 'images' in the storage config forCephFS? After importing the image you still can migrate it to the RBD pool. Or select the target storage in the import dialog.

Get multiple OSDs per node and 25G network. You are running the bare minimum for a working system.

Usually bonding is used with LACP and a corresponding switch stack to provide network redundancy. As Ceph uses multiple TCP connections both physical links are utilized.

It is compatible with RBD and qcow2.

Es muss kein aufwendiges shared Storage dranhängen, um den Fehler zu reproduzieren. Ich habe die notwendigen Schritte hier dokumentiert: http://gurubert.de/ocfs2_io_uring.html

ceph osd pool set $poolname crush_rule $rulename will assign a rule to an existing pool. Ceph will then move all the PGs with their objects according to the new rule.

Bei Linbit gibt es ein DRBD-Plugin für Proxmox. Das kann für so kleine Umgebungen besser sein.

Das ist mit aktuellem Ceph auch nicht mehr notwendig. Früher war das IO auf SSDs/NVMes noch nicht so optimal, so dass 2 OSDs auf einem Device empfohlen wurden. Wenn das Device dann aber defekt ist, sind halt auch beide OSDs kaputt.

Selbst wenn Du eine schnelle Netzwerkkarte besorgst, wirst Du mit diesem absoluten Minimalsetup nicht glücklich werden. Nur eine OSD pro Knoten und nur drei Knoten ist einfach zu wenig, Ceph kann damit seine Stärken nicht ausspielen. Die Performance wird auch lange nicht an die möglichen...

You could define the backup jobs to backup all VMs from a specific pool. Then add the new VM to that pool and it will be automatically backed up.

dd is a command line tool to write bytes from and to a file or block device.

You should add that you need to first get the OSD map with ceph osd getmap -o om.

/etc/pve is a FUSE mount that gets its data from the Proxmox database. Without a running Proxmox there is nothing in /etc/pve.

All nodes in the Proxmox cluster should run the same versions of the software.

It looks like the kernel builtin CephFS client is too old to talk to the MDS.

IMHO you need to choose the method of management for the Ceph cluster. It is either with the tools and the GUI of Proxmox or with the cephadm orchestrator, but not both at the same time. You can setup Ceph with cephadm alongside Proxmox on the same hardware and use it as storage for Proxmox.

`virtio` may be something completely different. Please switch to `scsi` and the `virtio-scsi-single` virtual HBA in the VM.

You could add a standby-replay MDS instance that will be active faster than a cold standby. Each CephFS needs at least one active MDS. With 8 CephFS you would need 8 MDS instances. Do you have the capacities to run these? Multiple MDS per CephFS are needed when the load is too high for just...