Search results

You already have a unhealthy cluster before you do the reboot. Status should be HEALTH_OK before you start any maintenance work. osdmap e845: 7 osds: 2 up, 2 in; 8 remapped pgs So you have 7 OSD's (HDD's) and only 2 of them are online/working!? Also 1 mon is down. You have 4 mons, so when you...

What is the status op CEPH when you shut down one node (ceph -s), is it really going into RO-mode? How many mon's do you have and how many OSD's per node? Can you post CEPH config and crush map?

Duplicate, please see https://forum.proxmox.com/threads/networking-problem-in-4-1-33.27035/ and https://bugzilla.proxmox.com/show_bug.cgi?id=952

No, you can't mix PVE 3.x and PVE 4.x. But if you use shared storage you can simply shutdown the VM on PVE 3.x and move the .conf file to a PVE 4.x node and start the VM. This way we migrated many VM's from PVE 3.x to 4.x without any problem and almost zero downtime.

Update: I just did a double reboot of each node and migrated the VM's the same way I did yesterday. After all the nodes were rebooted I migrated them back to the node where they belong. Everything was OK, to be sure I stop-start every VM one-by-one, also no problems. Strange, but can't reproduce...

No, this are the packages that were updated (maybe Ceph relies on one of these, but don't know that): iproute2: 4.2.0-2 ==> 4.4.0-1 libelf1: 0.159-4.2 (new) libnvpair1: 0.6.5-pve7~jessie ==> 0.6.5-pve9~jessie libpve-access-control: 4.0-13 ==> 4.0-16 libpve-common-perl: 4.0-54 ==> 4.0-59...

Okay, any other suggestions where to look for this problem? In logfiles nothing is found. Tonight i'm going to move all VM's and reboot each node again, just to test what happens. Something special to watch for?

On a small 3-node cluster, running 16 VM's, I did a upgrade from 4.1-22 to 4.2-2. Each node also has one CEPH OSD onboard. Cluster (Proxmox VE and CEPH) was completly healthy before the upgrade started. I did the upgrade node-by-node and before I finished a node and started with the following...

Since new GUI (PVE 4.2) a right click on a VM redirects you to the VM summary. With the old GUI this wasn't the case and you stayed where you are. I don't like this new behavior and hope it can be changed back? Thanks!

Can you post the output of "mount" and "df -h"?

All I'm saying is that a page/overview with the current versions in all repositories can be helpful and handy so you have all the information at a glance.

Is somewhere showed what version is currently the latest for each repository? So for example a webpage (or even on the PVE homepage?) with a block like: Latest releases: pve-enterprise: 4.1-22 no-subscription: 4.1-34 pve-test: 4.1-?? For exmple like FreeBSD (http://www.freebsd.org) does...

What are the devicenames of your CD/DVD drives (ide0, ide2)? Can you swap the images? So if ide0 is now used for VIRTIO drivers and ide2 for the Windows ISO, swap them and try again. What is your bootorder? If you only have the Windows ISO "mounted", will it boot? (you can't finish installation...

You still have no quorum. Test multicast traffic with omping.

In /etc/network/interfaces and /etc/hosts (and restart the nodes).

You simply have no quorum (and therefor activity is blocked and you can't use /etc/pve) it looks like you use your WAN IP for the Proxmox VE cluster? Are you sure your ISP allows multicast traffic? Did you test with omping? I have experience with Proxmox VE 3.x and 4.x, but I think PVE 4.x is...

I also like the new features in Jewel, but I really hope Proxmox VE wait at least until PVE 4.3 or later before it starts using Jewel. There's nothing wrong with a conservative policy for this kind of major updates. Especially within enterprise products like PVE is. If you are not on a...

Can't help you with this problem, but will follow this thread closely, since we are planning to use ZFS for our backups too (and for our VM storage already use CEPH).

Hmm, no I don't have any other clues how to enable nested support for you, sorry. You can disable KVM hardware virtualisation (options tab of the VM), so you can at least start the VM. But performance would not be very good. :(

Is there anything in the logs (messages, syslog and/or daemon.log) about fencing? Not sure why this occurs, but I guess the node is rebooted because the watchdog timer expired and thus the node is fenced. If my guess is correct, the only question left is why the node is that busy when a backup...