Search results

We have a small 3 node cluster where the Ceph monitor store is suddenly multitudes larger than it was previously. We typically 'systemctl restart ceph.target' when we observe Ceph packages having been updated and only schedule node restarts to apply newer kernels or Intel microcode updates...

Old thread, hopefully this is informational to others: Hard disk manufacturers correctly use the term Tera Bytes, eg 8TB which is 8,000,000,000,000 bytes. The storage term everyone else has miss used, is a multiple of 2 and called Tebi Bytes (TiB). 8 TB = 8,000,000,000,000 / 1024 / 1024 / 1024...

Herewith a refined version: #!/bin/bash time='60'; filter='rbd_hdd'; function getstats() { for host in `ls -1A /etc/pve/nodes`; do if [[ "$HOSTNAME" == "$host" ]]; then iostat -xkdy $time 1 | grep '^rbd' > /tmp/"$host"_iostat & else ( ssh -o StrictHostKeyChecking=no...

The key objectives around Ceph is for it to be an easily managed, reliable and scalable storage architecture. Replacing an OSD should be as simple as replacing the old drive and running a single command which then brings it in to service. Typical Ceph deployments have OSD counts in the hundreds...

We're running Ceph Luminous with latest updates and no longer observe these errors

Most systems processed updates without problems but we have one which exhibits the following. Is this possibly due to us being routed to an out of sync mirror or necessitate more careful investigation? [admin@kvm2 ~]# apt -o Acquire::http::AllowRedirect=false update Ign:1...

Memory and CPU utilisation is accurately displayed when reviewing PVE node summary. It would however be nice to sum up all individual CPU metrics as it shows the Hyper Threading performance loss (green area is shown to be shrinking in Zabbix). The load average relates visually but the values...

You should define NUMA regions to match the physical host's and draw CPUs from pools. Documentation is unfortunately very sparse (numaN): https://pve.proxmox.com/wiki/Manual:_qm.conf The following sample, most relevant somewhere like rc.local, would then need to run after starting the guest...

Interesting reading: https://blog.cloudflare.com/how-to-receive-a-million-packets/ https://blog.cloudflare.com/how-to-achieve-low-latency/amp/

You could try the following: Disable Hyper-Threading in the host's BIOS Pin the vCPUs to cores that are on the same processor the host NIC connects to. Reference the motherboard's block datagram There was talk that it's better to assign CPUs rather than cores to VMs but I assume that relates to...

Windows guests running stable VirtIO drivers incorrectly identify disc images as being thinly provisioned. This results in automated weekly defrag operations not consolidating days and instead attempting to pass through discards. Performance of SQL VMs on single Proxmox nodes (no Ceph) has...

Yip, still recommend that Proxmox update their wiki to get users to convert all straw buckets to straw2 and most deployments would benefit from minimising full OSDs due to uneven distribution of data. Would be nice if Proxmox updated monitoring tools to be Luminous based so that we could switch...

I've been using FreePBX as a GUI to manage Asterisk for a long time and really value the flexibility the developers provide by having hooks available pretty much everywhere. Where would I log an official feature request, to have Proxmox fire scripts when booting or shutting down VMs where...

Back on topic for this thread. This document (https://software.intel.com/en-us/articles/setting-up-intel-ethernet-flow-director) details that the Intel Flow Director will automatically program the NIC with hashes of outgoing flows so that return traffic is received back on the same CPU: Using...

MTU does need to be increased. Finding the maximum supported MTU requires one to step through various values until the command reports an error. We essentially start at the maximum allowed by our switches (ifconfig eth0 mtu 9216) and then step down until we don't get an error (Dell servers...

Spirit, perhaps you or someone else at Proxmox could edit the Wiki and documentation to provide a quick overview of the 4 levels of activation necessary to get this working? I personally also initially had problems understanding the multiple activation requirements and value the granular...

Got this working by defining an alias called 'multicast' with a CDIR of 224.0.0.0/4 Then defined the following rule: Applicable sections from /etc/pve/firewall/cluster.fw: [ALIASES] multicast 224.0.0.0/4 # Multicast ceph 10.254.1.0/24 # Ceph [RULES] IN Ceph(ACCEPT) -source ceph # Ceph -...

Proxmox firewall rules are automatically generated to cover cluster communications by discovering and defining the 'local_network' ipset: [root@kvm1a ~]# pve-firewall localnet local hostname: kvm1a local IP address: 192.168.241.2 network auto detect: 192.168.241.0/24 using detected...

Interesting, I'll read up on ifupdown2. My requirement is actually the exact opposite, I want frames ingressing a LACP bond on the Proxmox host to have the first VLAN tag (standard 802.1Q aka TPID 0x8100) stripped when it is presented to the virtual guest and to wrap anything and everything...

I did mention in my earlier post that I was going to follow Red Hat's official recommendation of avoiding OvS LACP bond interfaces by assembling a Linxu LACP bond and then attaching that to the OvS bridgef or VM traffic. I'll provide feedback on whether or not this addresses the throughput and...