Hi folks,
I've got a persistent issue on my new Proxmox instance - getting a kernel panic everytime when trying to test ZFS filesystem (not zvol) with recordsize 1M, 512K affected as well. Here is test case which I've used to reproduce - it worked in 100% runs and I even was able to reproduce it even on VMware virtual instance as well:
- install Proxmox from the latest official ISO;
- configure pve-no-subscription repo and update (apt-get update && apt-get -y dist-upgrade);
- configure ZFS arc limits (zfs_arc_max=6442450944 and zfs_arc_min=3221225472) in /etc/modprobe.d/zfs.conf;
- configure none IO scheduler for 8 external connected SATA HDD;
- create ZFS RAIDZ2 pool using Proxmox GUI (ashift=12 compression=off, name=testpool) using 8 external connected SATA HDD;
- set the following parameters for created ZFS dataset: zfs set atime=off relatime=on xattr=sa dnodesize=auto logbias=throughput sync=standard dedup=off recordsize=1M compression=off testpool;
- install bonnie++ package;
- make a dir for test, add a user (bonnie++ not allow to run a test under root) and set permissions for dir: useradd frank && cd /testpool && mkdir bon && chmod 777 bon;
- reboot host;
- run a basic bonnie++ test: bonnie++ -d /testpool/bon/ -x 1 -u frank
I'm getting a kernel panic constantly in this test case, recordsize 512K affected as well. While using recordsize 256K or lower this test runs without errors. According to dmesg dump which I've captured using kdump-tools it seems that unknown memory leak in ZFS is triggered somehow during test - I've attached it for analysis. Also I've tried the opt-in kernel (6.2), 5.11 kernel and with no luck - faced the same error. Could you pls guys help?
I've got a persistent issue on my new Proxmox instance - getting a kernel panic everytime when trying to test ZFS filesystem (not zvol) with recordsize 1M, 512K affected as well. Here is test case which I've used to reproduce - it worked in 100% runs and I even was able to reproduce it even on VMware virtual instance as well:
- install Proxmox from the latest official ISO;
- configure pve-no-subscription repo and update (apt-get update && apt-get -y dist-upgrade);
- configure ZFS arc limits (zfs_arc_max=6442450944 and zfs_arc_min=3221225472) in /etc/modprobe.d/zfs.conf;
- configure none IO scheduler for 8 external connected SATA HDD;
- create ZFS RAIDZ2 pool using Proxmox GUI (ashift=12 compression=off, name=testpool) using 8 external connected SATA HDD;
- set the following parameters for created ZFS dataset: zfs set atime=off relatime=on xattr=sa dnodesize=auto logbias=throughput sync=standard dedup=off recordsize=1M compression=off testpool;
- install bonnie++ package;
- make a dir for test, add a user (bonnie++ not allow to run a test under root) and set permissions for dir: useradd frank && cd /testpool && mkdir bon && chmod 777 bon;
- reboot host;
- run a basic bonnie++ test: bonnie++ -d /testpool/bon/ -x 1 -u frank
I'm getting a kernel panic constantly in this test case, recordsize 512K affected as well. While using recordsize 256K or lower this test runs without errors. According to dmesg dump which I've captured using kdump-tools it seems that unknown memory leak in ZFS is triggered somehow during test - I've attached it for analysis. Also I've tried the opt-in kernel (6.2), 5.11 kernel and with no luck - faced the same error. Could you pls guys help?