ZFS backups from 64-bits Ryzen desktop to 32-bits Pentium desktop

I spend Saturday to restore my Backup Server, a server that I use since June 2019. It stores my second backup, the first backup is on my laptop:) The main problems I had were:

  • the motherboard battery had to be replaced;
  • Contact problems with the IDE connectors;
  • Many errors on the 250GB HDD according to the OpenZFS file system.

The following picture is my backup server, the keyboard (PS/2) is a leftover from the time with the depleted battery. I had to press F1 to overrule the missing date/time in the BIOS. Anyhow I copy the time from the Internet during booting.

The system configuration is:

  • a ~1998 Compaq Evo Tower with a CD R/W, DVD ROM and Floppy Drive, these are not connected. The relative modern power supply does not fit into the intended location and is mounted by one screw and some tape.
  • the motherboard is from a 2003 HP d530 SFF, it has a Pentium 4 HT (1C2T; 3.0GHz);
  • 2 working 512MB DDR memory sticks (400MHz).
  • 4 HDDs (2x IDE 3.5" and 2x SATA-1 2.5"), see SW description.
  • the motherboard has 1Gbps Ethernet and many USB2.0 connectors.

Beside the keyboard the system has two cables connected Ethernet and Power :slight_smile:

The software I run is 32-bits FreeBSD 13.2 (the latest release from May 2023!) on OpenZFS and it receives the backup from my desktop with Ubuntu 22.04 LTS and OpenZFS by using the commands “send | ssh receive”. Nowadays we have perfect compatibility between Ubuntu/Linux and FreeBSD due to an integrated development project.

I have the zroot datapool, that contains the FreeBSD and all my backups. It is stored on Raid-0 3x 320GB HDDs all with different speeds. The fastest HDD is a Seagate laptop HDD, followed by the IDE HDD and followed by a Hitachi laptop HDD.

I also use the 250GB HDD again in a separate datapool, because I did not trust it, because of the problems of the past. Probably it is no HW issue, because all HDDs have between1 and 3 power-on years. That 250GB HDD seems to work again. For the moment I backup one dataset on the 250GB HDD and to be safe it is also backed up on the Raid-0 configuration. By the time I need the space because of the growing number of snapshot, I might delete that double backup from the Raid-0 datapool.
Up to now the 250GB HDD seems OK, since I did run “zpool scrub” without errors. That program recalculates the CRC from the file and compares it with value stored in the directory during writing.

While working on the PC, I detected that hyper-threading had been switched off in the BIOS, so I switched it on again. The result has been that my transfer rates improved from ~23MB/s to ~29MB/s. The limit is caused by a ~95% CPU load on one of the threads.
I also switched the boot from the 2 IDE disks to the 2 SATA-1 laptop disks to avoid contact problems during the boot processes.