Ok, sadly this is expected. In 6.8.12-29 there is a known regression in context with IOMMU which causes these types of errors. See [1] for more info.
In this case please install and pin kernel 6.8.12-28 as suggested in the other thread, which is...
Ok, sadly this is expected. In 6.8.12-29 there is a known regression in context with IOMMU which causes these types of errors. See [1] for more info.
In this case please install and pin kernel 6.8.12-28 as suggested in the other thread, which is...
This doesn't work, because pve-network-interface-pinning will not generate if there is a pinned name already. So, I'll have to remove the Links first. Now I'm not very confident in your answer, I'm sorry to say, as your way won't run in the first...
Thanks for providing the info!
As you can see here
Ceph actually knows about the traffic but doesn't seem to consider it for the status page.
While i can't say for certain if this will fix the issue, i'd still urge you to update your servers...
Yes.
- First run pve-network-interface-pinning as shown above
- Then, afterwards, remove the .link for the old interface name under /usr/local/lib/systemd/network/.
- Finally, reboot
The pve-network-interface-pinning takes care of updating all...
Hello torbho,
You are right after a reboot the OSDs go "down"... and the system restart before the OSDs go "out".
For the configuration "mon_osd_down_out_interval" we have the default 10mn.
So no "noout" necessary, thank you.
Francis
Yes, that's what I tried to communicate. So, the removal of these links is all I need? Then I can pin the names I want? It read in the posts out there like other configuration files had to be edited, too.
Hello torbho,
You are right after a reboot the OSDs go "down"... and the system restart before the OSDs go "out".
For the configuration "mon_osd_down_out_interval" we have the default 10mn.
So no "noout" necessary, thank you.
Francis
If they are named nic4 and nic5, then they are were already pinned by the installer. It's our default naming scheme.
Afterwards, you just need to remove the old file at /usr/local/lib/systemd/network/50-pmx-<oldname>.link.
In order to upgrade I had to re-apply the commands you provided on a previous post:
rm /run/proxmox-datacenter-manager/api-cache/remote-*/subscription-state.json &&\
proxmox-datacenter-manager-admin support-status update &&\
apt update...
Hello Francis,
I think there may be a misunderstanding.
With a normal reboot, setting noout is usually not required. When the OSDs on a node stop, they are first marked as down. They are only marked as out after the mon_osd_down_out_interval...
Hello,
it was a browser based issue. I cleared the brower cache & cookies for that specific ip address only and keyboard (and mouse) work again.
Thank you
An.drea
Hi @jtheisen.
At the start of the PVE there were a lot of DMAR errors (and before update, none of them were appearing):
dmesg:
Jun 11 20:33:41.216928 pve1 kernel: I/O error, dev dm-8, sector 0 op 0x0:(READ) flags 0x0 phys_seg 32 prio class 2...
Hi,
you can use the pve-network-interface-pinning tool: https://pve.proxmox.com/pve-docs/pve-admin-guide.html#_using_the_pve_network_interface_pinning_tool
Especially the section on pinning a specific interface should do the trick...
Hello Torbho,
Thank you, I want to execute "ceph osd set noout" only when I reboot a node, not when a node "crash".
The problem, some times the administrator do not execute the "ceph osd set noout" and after a node update "reboot" the node...
I setup a PVE 9 to include in my Cluster. It pins the NIC names to static names. I didn't pay enough attention, and now my 10g NIC for the storage are nic4 and nic5. I'd like them to have a name like fiber10g1 and fiber10g2
Now, there are lot of...