Add NFS crashed, XFS and deprecate ZFS mover.

Work continues; work around bugs found via use of a different filesystem for cache.

README.md

@@ -25,7 +25,8 @@ The following open-source projects seem to be able to help reach my goals. It re
 
 - [SnapRAID](https://www.snapraid.it). Provides data parity, backups, checksumming of existing backups. 
     - [Claims to be better than UNRAID's](https://www.snapraid.it/compare) own parity system with the ability to 'fix silent errors' and 'verify file integrity' among others.
-- [BRTFS Filesystem](https://btrfs.wiki.kernel.org/index.php/Main_Page) similar to ZFS in that it provides be the ability to 'send/receive' data streams (ala `zfs send`) with the added benefit that I can run individual `disk scrubs` to detect hardware issues that require me to restore from snapraid parity. 
+- [BRTFS Filesystem](https://btrfs.wiki.kernel.org/index.php/Main_Page) similar to ZFS in that it provides be the ability to 'send/receive' data streams (ala `zfs send`) with the added benefit that I can run individual `disk scrubs` to detect hardware issues that require me to restore from snapraid parity. **My observed Btrfs performance is that its poor compared to XFS filesystem on linux.** *Since we use btrfs only for the 'data' disks in the slow mergerfs pool we are not sensitive to speed.*
+- **XFS Filesystem for NVME cache on mdadm array**. After finding bugs and instability in my ZFS+NFS+mergerfs implementation my cache disks are now formatted to XFS in RAID1. I did not use btrfs raid1 natively here because btrfs performance was poor (50% throughtput penalty). XFS was able to match ZFS raw speeds (without arc) ~900MB/s.
 - [MergerFS](https://github.com/trapexit/mergerfs). FUSE filesystem that allows me to 'stitch together' multiple hard drives with different mountpoints and takes care of directing I/O operations based on a set of rules/criteria/policies.
 - [snapraid-btrfs](https://github.com/automorphism88/snapraid-btrfs). Automation and helper script for BRTFS based snapraid configurations. Using BRTFS snapshots as the data source for running 'snapraid sync' allows me to continue using my system 24/7 without data corruption risks or downtime when I want to build my parity/snapraid backups.
 - [snapraid-btrfs-runner](https://github.com/fmoledina/snapraid-btrfs-runner). Helper script that runs `snapraid-btrfs` sending its output to the console, a log file and via email. 
@@ -38,14 +39,27 @@ The following open-source projects seem to be able to help reach my goals. It re
 apt-get install zfsutils-linux cockpit-pcp btrfs-progs libbtrfsutil1 btrfs-compsize duc smartmontools
 ```
 
-## ZFS cache pool setup
+## ~~ZFS cache pool setup~~
+**WARNING! DEPRECATED** NFS+ZFS is unstable with this setup. Follow XFS+mdadm below.
 
 RAID1 of two SSD disks. We'll write all stuff here then purge to 'cold-storage' slower disks via cron.
 
 ```
 zpool create -o ashift=12 cache mirror /dev/sdb /dev/nvme0n1
 ```
 
+## XFS RAID1 mirror mdadm
+
+See [mergerfs](mergerfs.md) for details on ZFS instability. For our cache pool we will use XFS filesystem. Set up the NVME cache as follows:
+
+```
+mdadm --create --verbose /dev/md0  --bitmap=none --level=mirror --raid-devices=2 /dev/nvme0n1 /dev/sdb
+mkfs.xfs -f -L cache /dev/md0
+mdadm --detail /dev/md0
+```
+
+Remember to add a mountpoint to start at boot. 
+
 ## BTRFS (disk setup guide)
 
 ### BTRFS Commands TL;DR

filemover/uncache-mover.py → filemover/zfs-uncache-mover.py

@@ -1,7 +1,8 @@
 #!/usr/bin/python3
 # TheLinuxGuy ZFS cache pool mergerfs tiered cache mover.
 # File age time-based mover depending on goal % cache utilization.
-# TODO: move logs to standalone log file to stop adding crap to syslog
+# This script works but is abandoned after NFS+ZFS+mergerfs instability.
+# !! THIS SCRIPT IS ZFS POOL SPECIFIC !! DO NOT USE ON XFS cache setup.
 import argparse
 import subprocess
 import syslog

mergerfs.md

@@ -1,5 +1,7 @@
 # MergerFS 
 
+**WARNING: Using ZFS + NFS (non-zfs native export) + mergerfs cause [ZFS mount instability and crashes](https://github.com/trapexit/mergerfs/discussions/1098).**
+
 MergerFS is used to "merge" all physical distint disk partitions (/mnt/disk*) into a single logical volume mount.
 
 ### Policies
@@ -44,6 +46,58 @@ To attempt to mirror what unraid provides with their share "cache" we are going
 
 Recall that I chose to use ZFS and RAID1 mirror for this purpose to provide assurances that my data would not be lost before it gets moved onto parity-protected-snapraid-slow-storage-disks.
 
+## NFS instability
+
+`/mnt/cached` is my mergerfs pool and ZFS mountpoint on my local system. The `mergerfs` process seems to be crashing at some point due to NFS. I haven't yet found the root cause of this issue and have tried everything from upgrading kernel, ZFS, nfs-kernel-server, libfuse and OS (Ubuntu 20.04 to 20.10).
+
+The crashes seem to be more pronounced when using NFSv4 protocols. NFSv3 is more stable but that is a stateless protocol and I would much prefer v4 only NFS shares. I have disabled v4 and force v3 for the time being to try to make my implementation stable. 
+
+Observed behavior (on local NAS):
+```
+# ls -lah /mnt/cached
+ls: cannot access '/mnt/cached': Input/output error
+```
+
+Recovery steps
+
+
+### Debugging with strace
+
+```
+root@nas:/home/gfm# strace -fvTtt -s 256 -p PIDHERE -o /tmp/mergerfs.strace.txt
+strace: Could not attach to process. If your uid matches the uid of the target process, check the setting of /proc/sys/kernel/yama/ptrace_scope, or try again as the root user. For more details, see /etc/sysctl.d/10-ptrace.conf: Operation not permitted
+strace: attach: ptrace(PTRACE_SEIZE, 2081428): Operation not permitted
+root@nas:/home/gfm# echo "0"|sudo tee /proc/sys/kernel/yama/ptrace_scope
+0
+``
+
+If that doesn't work, change setting `/etc/sysctl.d/10-ptrace.conf` to 0. Reboot.
+
+Strace isn't helpful according to mergerfs developer. Here's the proper way to debug mergerfs using gdb
+
+### gdb debugging mergerfs
+
+```
+If it's crashing then strace is pretty useless. Need a stack trace from gdb.
+
+gdb path/to/mergerfs
+
+run -f -o options branches mountpoint
+
+when it crashes
+
+thread apply all bt
+```
+
+### Remove ZFS from the equation by using XFS RAID 1
+
+```
+mdadm --create --verbose /dev/md0  --bitmap=none --level=mirror --raid-devices=2 /dev/nvme0n1 /dev/sdb
+mkfs.xfs -f -L cache /dev/md0
+mdadm --detail /dev/md0
+```
+
+
 ### NFS tweaks that were added
 
 ```