fresh installs of 9.3 are going straight to 4gb in first hour and crashing before I can even restore a backup
I have tried a good dozen fresh re-installs of 9.3 today thinking I might have to roll back to the 8’s and hope the cores are not updated edit
crap I just missed out on doing a 8 install and now they updated it for 10 dev and made 9 stable :S
gave 10-dev a spin and it died before it even created a vm
tho I think most of the issue is google-drive-backup
as it seems to be installed first then tries doing a backup while the system is being restored …
When I was playing around with my bluetooth dongle trying to make it work with dbus my memory usage just exploded. My server nearly used all memory but I was able to kill “node” process eating nearly all memory. I don’t know what was node process is/was that was eating all the memory. I don’t have ONVIF integration.
This could be related to dbus but this is just my guess.
And this is how it looks when it start to eat memory. I can’t find out what this node proccess actually is?
I had hassos running on a pi4 with 8gb Ram that kept hanging until a switched it off and on again, migrated over to a VM on my QNAP and spotted on vnx an out of memory error.
Watching the hardware memory usage graph the memory would climb every minute until it maxed out and core restarted or I had to reboot.
Disabled just about everything when I found this post and I can confirm it was ONVIF. I disabled my 2 ONVIF cameras and the problem stopped immediately.
Home Assistant 2022.11.5
Supervisor 2022.11.2
Operating System 9.3
Frontend 20221108.0 - latest
I still have the issue - and none of the solutions in this thread have proven a root cause. Look for a better way to troubleshoot, but my HA instance is running out of memory about every 3 days…
I ended up moving all my databases to be external ,
seems to now take 3 days to reach 3Gb instead of minutes , still no idea what is using the Memory
The only solution I can offer is do a backup and restore using a old version
I will just not run any updates until the dev’s stop pointing fingers at someone else
I had noticed memory usage double 1 month before this massive memory leak
I am very happy to read this thread, cause I was read to drastically upgrade my hw, since I associated the memory leak to its age
Unfortunately, I have the same issues of you. I don’t use the ONVIF integration and I don’t have any kind of camera installed.
I already tried to stop integrations to understand if some of them were making the ram usage grow so fast, but none of them seemed to cause the issue.
I don’t thing is related to DB usage, as I only track few entities (3 to 5 usually) for a timespan of 3 days and I regularly purge it.
I think I will go for the restart automation or for the hack suggested by @Neil_Brownlee, even if it is only a problem walkaround and does not actually solve its root cause.
I would like to understand the root cause, to definitely solve it.
I seem to have at least identified the big memory usage jumps. Every time a Home Assistant Core update is installed (including making a backup) and the core is restarted, my memory usage more than doubles. The only way to solve it (for me) is to restart the host, after which the memory usage drops back to about 1,2GB. Since I’m running Home Assistant OS I do this from the CLI.
Who else is experiencing this?
Note that I’m also experiencing the very slow increase in memory usage as well (about 200MB in the last 3 days).
I’m running Proxmox (7.3) on a NUC (i3 with 32GB memory and an SSD) with a VM running HAOS and another VM running Deconz (on Ubuntu). And I’ve got the hack suggested by @Neil_Brownlee running as well (which kills rogue HA instances).
I’m having a similar issue. Every time I run a backup on Home Assistant, memory usage climbs to ~90%. Did you ever figure out the issue? I’ve had this problem for several weeks and only recently connected it with a backup running.
having same issue. Had Memory-Profiler running and saw that the “camera” integration (shipped with HassIO) was taking over 1GB of RAM in short time. So i’ve disabled the old camera integration and moved to the new integration.
But the leak is not gone completely. Now the memory takes longer to reach 100%.
I notice the slow uptick in memory use on my system. I’m also seeing a very sharp uptick in memory usage (takes 2-3 minutes to go from 2 GB to just under 4 GB) when I run a backup. Does a similar thing happen on your system as well?
For me, leakage was noticed by days, not as you mentioned by hours.
I catch mine after migration to the new version of HA.
But curious, when I rollback the version of HA memory leak disappears.
For anyone who’s curious, I think figured out my issue. For some reason (that may be particular to my setup). Proxmox is displaying the sum of my used and buffered/cached memory, not just my used memory.
What appears to be happening is this. Every time I run a back up, my used + buffered/cached memory essentially maxes out. After the backup, the buffer/cache isn’t purged, so Proxmox reports my memory being almost maxed out.
I verified this by doing the following:
Using the directions on this page, I enabled root ssh access to Home Assistant OS.
After running a back up, I ran free -h, a Linux command to output a summary of my memory usage. It reports the following:
# free -h
total used free shared buff/cache available
Mem: 3.8Gi 1.2Gi 98Mi 4.0Mi 2.5Gi 2.6Gi
Swap: 1.3Gi 0.0Ki 1.3Gi
I run the following command sync; echo 3 > /proc/sys/vm/drop_caches. This clears the pagecache, dentries, and inodes. See this page.
Then, I run free -h again. It reports the following:
# free -h
total used free shared buff/cache available
Mem: 3.8Gi 1.2Gi 2.2Gi 4.0Mi 424Mi 2.6Gi
Swap: 1.3Gi 0.0Ki 1.3Gi
Here’s what Proxmox reports at the same time. (The dip is the moment I clear my buffer/caches.)
Why Proxmox displays things this way (or why my setup works this way) makes no sense to me. But at least I now understand why Proxmox reports such high memory usage but Home Assistant
seems fine.
