Search results

If ceph -s is timing out, and the network works, and the firewall is disabled (Datacenter -> Firewall -> Options), then check the state the MONs are in. The Ceph documentation explains how...

Check if the network configured for Ceph is working. Did you enable the firewall without configuring exceptions for Ceph (macro)? If you use a large MTU, make sure to use ping with a large payload to fill up the packet so that in the end with the ICMP and IP overhead, it will be the size of the...

Ich hab jetzt nicht den ganzen Thread durchgelesen, aber vielleicht hiflt der Hinweis: iperf3 in den Versionen wie man sie mit Proxmox VE 8 / Debian 12 bekommt, ist nicht multithreaded! Aus man iperf3: Und wenn es um schnelle Netze geht, muss man da mitunter mit mehr als einem Thread testen...

https://docs.ceph.com/en/quincy/architecture/#smart-daemons-enable-hyperscale The diagram at the end of this section shows the flow of data. The network diagram might also help: https://docs.ceph.com/en/reef/rados/configuration/network-config-ref/ Client (VMs) to the primary OSD will go over...

I assume it will work, but haven't tested this myself! Looking at the date of the article and the history of our full-mesh article, it appears that it was written just a few months before we added the "routed with fallback" option, which is based on OpenFabric.

We have some customers who set up a 5-node full mesh with the OpenFabric variant by using 4-port 25 Gbit NICs. But those were all full Ceph nodes AFAIR. Seems to be working quite well for them and up to 5 nodes is still manageable regarding the wiring. ;)

Do you see anything else in the journal of the OSD? The status output only has the last few lines. How are the permissions of the /etc/pve/ceph.conf file? Should be: -rw-r----- 1 root www-data 849 Nov 12 12:41 /etc/pve/ceph.conf With owner=root and group=www-data

Hmm, usually we also have the cluster_network line in the global section. It can be the same IP subnet as the public. Make sure that the network is configured onthe hosts and that they can reach each other via this network. The simplest way to change this without any downtime would be to...

yeah, in my experience, defining the VM disks as SATA works if you install ESXi.

So, to clean this up on the Proxmox VE side, remove the node from the cluster: https://pve.proxmox.com/pve-docs/pve-admin-guide.html#_remove_a_cluster_node For Ceph, remove the missing services so that Ceph forgets about them. 2 / 3 MONs are still present right? ceph mon remove {mon-id} ceph...

Have you verified that the large MTU works? ping -M do -s 8972 {target host} The -s parameter is set to 9000 minus the 28 bytes of ICMP of IPv4 overhead.

The two main cases where the dirty bitmap is currently lost: - VM had seen a full stop. This is also the case if you use the "Reboot" option from the Proxmox VE UI to apply pending changes - The VM is backed up to more than one Backup Server, as the dirty bitmap is bound to a target Backup Server.

The license of Proxmox VE is always the same, primarily the AGPv3. We do sell subscriptions ;) There are no differences in limitations or features, no matter if a subscription is applied or not. Once you use it in production, subscriptions are recommended, as they give you access to the...

If you got plenty of RAM and don't seem to be running out of it, I would say, don't bother with SWAP at all :)

You mean the "AVAIL" column? This happens if some disk images, or in ZFS terminology, ZVOL datasets have a reservation / refreservation set. If the Proxmox VE storage config for it doesn't have the "thin provision" checkbox set, the datasets are created "thick" which for ZFS means, they get a...

NAME USED AVAIL REFER MOUNTPOINT […] PVE02_RAIDZ/vm-355-disk-0 16.0T 7.75T 13.4T - There are the 17T. Since this is a raidz, can you also show the zpool status PVE02_RAIDZ output? And the volblocksize of the disk image is of interest: zfs get...

Whats the output of zfs list? Ideally copy it into the post inside of blocks. Or use the </> Icon at the top to paste the output into.

There is no option to do it on the fly. You will have to free it up, destroy the LVM (and the storage config in Datacenter->Storage), then create the ZFS pool on the empty disk.

Zum MGMT von vielen einzelnen Instanzen (und live migration dazwischen) gibt es den PDM. Aktuell noch eine Alpha, aber ihr könnt sehen, wo die Reise hingeht. Zu den Backups: hier kommt es drauf an ob die Datenbanken auf Windows laufen. Denn da kann es in den Defaulteinstellungen zu Problemen...

okay, the other thing is, with whereis, you might need to specify that you look for the binary: whereis -b ceph-mon. Personally I use the which tool for this. another question I have is, how do you create a pool? Are you using the CLI or GUI? If you use the CLI, do you use the Ceph native tools...