B1659 Roon 2.67 : RAM usage feedback

[simon_pepper]

Building on the previous RAM usage analysis, there was the release of B1659 yesterday and Memory usage in Roon Server appears to be very controlled now

Sp post restart it was 3GB, and then with ‘Scheduled Activity’ increased to 5GB (where it was pre-update) and has remained.

[Wade_Oram]

Have you observed any increase in utilized memory (possibly a leak) around the time that backups are run. I previously commented on a significant increase in memory utilisation over the backup period:

Roon Backup appears to trigger a memory leak

Using the netdata systemd memory chart and filtering for just RoonServer, this is what I see over the last two days: [image] You can see two instances of additional memory being used at 4 AM with some (but not all) released about 3 hours later. My Roon memory usage does increase with time and/or Roon use but it does not increase to the point that I have to restart the service (or the server). This may be because of the limited size of my library. Note: my Roon usage over the period shown ha…

I am currently seeing much the same, with respect to the increase in memory during backups, on the production 2.66.

[simon_pepper]

My last Backup ran on the 12th May at a scheduled time of 3am, which is now in the middle of the later ‘Scheduled Activity’ window of 2am to 6am, so any memory usage involved in running a Backup is getting lost in the update activity.
I have now moved the ‘every 4 days’ Backup activity out to 6am, at which time all the library updates, metadata refresh, etc. has been completed.
Next scheduled back up is then the 16th May - so will look again.

[simon_pepper]

The 5GB usage has now gone to 6GB, following the ‘Scheduled Activities’ from 2am to 5am

The growth being in both managed and unmanaged memory

With 16GB available there is still 10GB of headroom on the server.

[simon_pepper]

There was a reboot like event with my ROCK server at 04:00 this morning.
The admin page says it been up for nearly 3 days, but there was a crash of some sort

Memory usage was down to 4GB from over 7GB.
There was a backup taken at 6am.

I wonder what happened?
Much of the unmanaged memory was dropped

ChatGPT tells me

The actual big memory reset event

The dramatic memory collapse in your graph appears later — around 05:00–05:05 UTC / 06:00 local — where:

• Physical memory suddenly dropped from ~6.3GB → ~4.1GB
• Handle count collapsed from ~975 → ~411
• Virtual memory also dropped sharply

Example immediately after reset-like behavior:

38805mb Virtual, 4109mb Physical, 2550mb Managed, 1559mb estimated Unmanaged, 411 Handles

That pattern strongly suggests one of:

• a process restart,
• a subsystem restart,
• or a major allocator purge/cleanup.

The huge drop in handle count is the strongest indicator this was more than just GC reclaiming memory.

And that’s when the Backup was scheduled for.

[simon_pepper]

Since last night memory event, there was gain in memory usage of over 2GB at the start of the “Scheduled Activity”

Current RAM usage is at about 5.5GB
And the jump was larged unmanaged