PBS tape backup how to improve data streaming performance

[AmateurProgrammer]

I have a setup with PBS backing up VM:s from PVE. Those backups are written to tape weekly.

I'm upgrading the tape drive from LTO4 to LTO5 but my setup has been struggling to keep up the streaming for the LTO4 drive with periodical shoe-shining occurring while running the backup jobs. As LTO5 has higher bandwidth I'm assuming I would need to upgrade some parts on my setup to prevent shoe-shining from happening in order to make the tape drive and tapes last longer.

My current setup is as follows:
Zpool:

NAME                        STATE     READ WRITE CKSUM
        backup-pool                 ONLINE       0     0     0
          mirror-0                  ONLINE       0     0     0
            wwn-0x5000cca27d0d5a04  ONLINE       0     0     0
            wwn-0x5000cca27d0b9818  ONLINE       0     0     0
        special
          mirror-1                  ONLINE       0     0     0
            wwn-0x55cd2e404c4d5612  ONLINE       0     0     0
            wwn-0x55cd2e404c4d614d  ONLINE       0     0     0

With the special vdev consisting of 120gb Intel S3500 series SSD:s and data disks 8TB 7200RPM HGST SAS3 drives. Pool has atime=off which considerably improved the read performance from the pool when running the tape backup jobs. The server currently has 16GB of ram.

Any suggestions on how to improve the performance while writing to tape? Would doubling the ram to 32GB have any positive performance effects on large-ish tape backup jobs?

 

[AmateurProgrammer]

Okay, it seems the problem was mostly with hba.

However I'm still experiencing weird behaviour: the io from the zfs pool seems to stop for a while, the tape drive stops and after few seconds the procedure continues again at full speed. My drives seem to deliver the constant 140mb/sec required to satisfy the maximum bandwidth of my LTO5 drive and I doubt that this is issue anymore with my zfs pool not providing data fast enough.

It looks to me that the tape backup software simply stops asking the zfs pool for data. I have checked the pool io using zpool iostatand the read speed seems to match mostly with the tape backup job reported ones, until it just abruptly stops. How could this be troubleshooted further?