because now I'm out of money, what is the exact value of 32G to put in the file /etc/modprobe.d/zfs.conf
cluster performance degradation
- Thread starter toto-ets
- Start date
From one of my nodes, just adjust:
To calculate such numbers by yourself:
Code:
~# cat /etc/modprobe.d/zfs.conf
#
# 4 GiB = 4294967296
# 8 GiB = 8589934592
# 12 GiB = 12884901888
# 16 GiB = 17179869184
# 24 BiB = 25769803776
# 32 GiB = 34359738368
#
# Nicht vergessen:
# update-initramfs -u
options zfs zfs_arc_min=8589934592
options zfs zfs_arc_max=17179869184To calculate such numbers by yourself:
Code:
$ echo "32 * 1024 * 1024 * 1024" | bc
34359738368in my /etc/modprobe.d/zfs.conf file I only find
options zfs zfs_arc_max= ...... I also have to insert options zfs zfs_arc_min
options zfs zfs_arc_max= ...... I also have to insert options zfs zfs_arc_min
You do not need to; my file is just an example
The default is described this: https://openzfs.github.io/openzfs-docs/Performance and Tuning/Module Parameters.html#zfs-arc-min
updates, I reinstalled everything with local ZFS Raid10 it seems that now the latency is better and the vms respond quickly, but I have a problem on the cluster, is it normal that if I restart a node the other one also restarts? currently the cluster is made up of 2 nodes, but I thought since now since the vms run locally it could also be done with just one node, why if I restart pve1 does pve2 also restart and all the vms go down?
yes I knew this, I'm preparing the 3rd node, but I wonder why pve2 also restarts if I reboot only on number 1?
I had another problem, I made a zfs pool with Raid 10 and 6 HDD, I tried to remove an HDD and the pool went SUSPEND, AND THE VMs ARE NO LONGER ONLINE, but shouldn't everything have continued to work with only one disk OUT?
How do I send the zpool clear command?
How do I send the zpool clear command?
Last edited:
from Shell
zpool scrub local_ZFSit should , try again
BTW, HDD are slow, even more with CEPH or ZFS
cannot scrub local_ZFS: currently scrubbing; use 'zpool scrub -s' to cancel scrub
with zfs raid 10 local they are faster
at the end of the scrub errors are cleared. during scrub , disks are slowdown , because all data are read to verify integrity.
at the beginning. Will be too slow for example for Windows as RDS.
no, I haven't tried CEPH yet.
but CEPH or ZFS kill HDD , because they add overhead ( where Windows is slow on HDD even on bare metal )
but CEPH or ZFS kill HDD , because they add overhead ( where Windows is slow on HDD even on bare metal )
Last edited:
I wanted to understand why ZFS, if I remove a disk, the pool is no longer reachable and it gives me suspended. Shouldn't a raid 10 with 6 HDDs also work with 3 less disks?
I can't buy sdd now, I bought 12 16TB HDDs, unfortunately they gave me the wrong advice
I can tell you that with ZFS locally in raid 10 we went much better than before I had ceph, and the 10G nics
Last edited: