Chasing nightly backup memory issue since 2026.6.x update

1

Running HA on Proxmox VM with 10G allocated memory, 250G SSD on a Beelink S12 Pro Mini. No issues until updating to 2026.6.x. Using the Mariadb in Apps. Have a Nabu Casa subscription. Run Synology native core integration. Nightly backups in 3 locations: Nabu, local and NAS. Backups around 1.65G in size.

Since update, I noticed the frontend crashes every night around 4AM, right when the database purge occured. Here is a rundown of my testing and observations.

I track the Open file descriptors python3 entity, which has remained steady at 134 to 150 consistently. So I ruled out integration leaking memory, which is the most likely culprit.

I set the following in configuration:

recorder:
  db_url: !secret db_url #This is my mariadb from Apps
  purge_keep_days: 5 #Lowers the database size

I moved the scheduled backup to 2AM while tracking memory on Proxmox. Memory went from 4.65G at 2AM to 9.15G at 2:30AM. With these changes my frontend didn't crash, but the memory didn't settle back down. I'm going to eliminate the Nabu Casa backup tonight to see if that is leaking memory.

Anyone else seeing memory issues similarly?

pve-Proxmox-Virtual-Environment-06-16-2026_09_41_AM
pve-Proxmox-Virtual-Environment-06-16-2026_09_41_AM566×592 13.2 KB

2

Are you running snapshot backups in proxmox?
If so, this locks up the db in HA, and your write ahead file (.wa1) might get bloated and will cause backups to fail - I have experienced this in my own system.

My solution was to disable short cycle snapshot for HA VM, and only do a weekly Snapshot. After some days, the .wa1 file is sorted and written to the main db and backups will work fine again.