Search results

This is not possible as the CephX keys for the OSDs and all the other internal config will be missing. What happened to your cluster? If you still have all the OSDs you could try to restore the cluster map from the copies stored there and start a new MON with that. Read the Ceph documentation...

Das WAL liegt üblicherweise zusammen mit der RocksDB auf einem Device. Es ist sehr unüblich, ein separates WAL-Device zu konfigurieren. Ceph nutzt immer Block-Devices und richtet sich dieses selber ein.

Das ist eine sehr alte Empfehlung aus den Anfangstagen von BlueStore. Inzwischen ist die RocksDB auch flexibler geworden. Für RBD und CephFS Nutzung reichen 40 - 70 GB, wenn S3 dazukommt, dann dürfen es auch 300GB werden. Der RadosGW nutzt viele OMAP-Daten.

The Proxmox installer is not built for that use case. You could install Debian alongside Windows, make it dual-boot and then install Proxmox on Debian.

Thanks for the mention. We only see OCFS2 as a transition technology for the case when the hardware is already there. OCFS2 has its issues. We would always recommend a supported storage setup for new infrastructure.

Why don't you run TrueNAS directly on the Hardware? What is the benefit of virtualization when you need to pass 21 drives into the VM?

You could assign the storage for the VMs only to the nodes where they should run. No storage, no images, no VMs.

A hotfix release 18.2.7 is currently being prepared by the Ceph project.

Do not run Ceph on 1G network equipment. You will be disappointed.

proxmox-kernel-6.14 ist already available for Proxmox 8.4

Yes, that is safe.

Das sieht doch gut ausbalanciert aus. Eine Empfehlung noch: Verdoppele die Anzahl der PGs für den Pool IKT-Labor. Momentan sind es weniger als 70 PGs pro OSD. SSDs können gerne 200 - 400 PGs pro OSD halten. Dann kann der Algorithmus die Daten noch besser verteilen, weil die einzelnen PGs nicht...

"ceph osd df tree" wäre interessant und "ceph df".

Thos cannot be avoided. If Ceph needs more space for the RocksDB it takes it from the block device.

You cannot just change the IPs in the ceph.conf file. You need to deploy new MONs first that listen to the new IP addresses. These addresses are recorded in the mon map as you already have seen. The IPs in ceph.conf are just for all the other processes talking to the Ceph cluster.

For 4. you need OCFS2. It can be used by Proxmax as shared directory storage like NFS. With qcow2 for the VM images you get thin provisioning and snapshots. But this setup is not officially supported by the company that makes Proxmox. You have to configure OCFS2 yourself.

With a SAN and LUNs presented to each Proxmox host you do not use Ceph as the SAN already has its internal replication (RAID). You can use LVM on top of the LUN as written in the Proxmox documentation. Or if you want to have something similar to VMFS you can use a cluster filesystem like OCFS2...

You have a LUN from a SAN presented to all Proxmox nodes via FC or iSCSI? This is something that cannot be used with Ceph. Ceph uses local disks and replicates over the network. With a LUN you can use LVM as storage for the VMs.

This is normal. Ceph adds the size of the RocksDB device to the total size of the OSD. But since the RocksDB device cannot store any data it is computed as completely used. This is why you see a 10% usage on your fresh OSD. BTW: 80GB RocksDB device seems a little bit large for a 750GB HDD. And...

To create client keys for mounting CephFS use the command "ceph fs authorize". It will create the necessary capabilities for the key.