Last days I decided to improve my experimental CEPH cluster (4 x PVE = 4 x OSD = 4 x 2TB HDD) performance by adding DB on small partition of NVMe.
To do this I need to cut some space from existing NVMe L2ARC partition.
Every PVE host has 2 x HDD for rpool, and rpool's ZIL and rpool's L2ARC are located on partitions on NVMe.
Everything went smoothly on 2 machines, but on 3rd machine after issuing command
With 2 previous machines it works smoothly, without any hang or lag.
I still can ping 3rd machine. Command
I don't think there is kernel panic, because I've added to every PVEs
I will inform when I got more information. There is similar issue reported to open-zfs project: removing L2ARC device from pool sometimes causes a freeze
To do this I need to cut some space from existing NVMe L2ARC partition.
Every PVE host has 2 x HDD for rpool, and rpool's ZIL and rpool's L2ARC are located on partitions on NVMe.
Everything went smoothly on 2 machines, but on 3rd machine after issuing command
zpool remove rpool /dev/nvme0n1p3 machine "hangs". With 2 previous machines it works smoothly, without any hang or lag.
I still can ping 3rd machine. Command
nmap returns some PVE normal open ports, but there is response from aby service from this machine.I don't think there is kernel panic, because I've added to every PVEs
panic=30 to kernel cmdline.I will inform when I got more information. There is similar issue reported to open-zfs project: removing L2ARC device from pool sometimes causes a freeze