[6.8] Observations around performance decrease

I don't know if this is important or already known. With a little change my problems went away this night.

I do have a lengthy backup job that runs every night. This job calls a lot of cp/rsync/stop docker/start docker/... commands. This job usually took ~2 hours on weekday, ~4 hours on Sundays and ~6 hours at the first day of a month. With 6.8 this changed heavily. The weekday job took around ~10 hours now. The weekly job never came back - I had to kill running rsyncs and the surounding job.

I could see Kbit/s values that got tranfered. The harddisks in use showed sporadic blips of the activity LED.

Here's a sniplet what the weekday job does. I show the Plex part only:

...
docker stop plex
rsync -avPX --delete-during --exclude Cache/ "/mnt/cache/system/appdata/plex/" "/mnt/user/Daten/unRAID/Tower/appdata_backup/plex/"
cp /mnt/cache/system/appdata/plex/Library/Application\ Support/Plex\ Media\ Server/Preferences.xml /mnt/user/Daten/unRAID/Tower/plex_backup/
cp /mnt/cache/system/appdata/plex/Library/Application\ Support/Plex\ Media\ Server/Plug-in\ Support/Databases/com.plexapp.plugins.library.db /mnt/cache/system/appdata/SQLite/Plex/
echo ".dump metadata_item_settings" | sqlite3 /mnt/cache/system/appdata/plex/Library/Application\ Support/Plex\ Media\ Server/Plug-in\ Support/Databases/com.plexapp.plugins.library.db | grep -v TABLE | grep -v INDEX > /mnt/user/Daten/unRAID/Tower/appdata_backup/plex/settings.sql
echo ".dump metadata_item_settings" | sqlite3 /mnt/cache/system/appdata/plex/Library/Application\ Support/Plex\ Media\ Server/Plug-in\ Support/Databases/com.plexapp.plugins.library.db | grep -v TABLE | grep -v INDEX > /mnt/user/Daten/unRAID/Tower/plex_backup/settings.sql
docker start plex
...

The weekly job transfers Backup data to another server. Here's an example. This command never came back:

...
rsync -avPX --delete-during --protect-args -e ssh "/mnt/user/Daten/" "[email protected]:/mnt/user/Backup/Daten/"
...

Yesterday I changed all occurences of /mnt/user to /mnt/disk17 (source and target of rsync) and booom. This night I was back to the old performance.

This is funny because in the past I never used User Shares. disk17 was the backup disk on all servers. Recently one backup folder on a remote server had to spread two disks. So I switched all backup jobs from Disk Shares to User Shares. This was working fine with 6.7.2. It became a problem with 6.8.

Just my 0.02 USD.