Over time my system appears to be getting slower and slower at start up. I realize I am making increasingly complex and rich in devices but some of the long entries in the snapshots below just seem unusually long and have little or nothing to do with my devices. Sun for example… >43 seconds? Moon > 42s Pi-Hole which just talks to a VM based Pi-Hole >40s?
I guess I don’t fully understand the “cost” of adding all these integrations. Also, it is unclear whether these times are all, or in part, concurrent. I haven’t clocked a restart but I believe it is quite a bit shorter than the total sum of the startup times listed…
It depends on the integration. Most are concurrent but some have dependencies that must be set up first.
The time is also dependant on what hardware you are running home assistant on.
If you look at the toast messages that pop-up (like toast) in the bottom left of your dashboard as home assistant starts you can see what the hold-up is.
My home assistant takes 70 seconds to start on a reasonably powerful core i7 processor. 60 seconds of this is due to the 3rd party Dahua integration. Dev has said nothing can be done
I have been monitoring start times and they are lower than the ones from the first post. My guess is that one of my Proxmox servers had not fully booted up and possibly delayed HA boot process due to services being down (ie DNS, etc). Either way I still have some that seem to take a long time but the fact that the number of seconds next to each integration is not how long it took to run, but just a marker of when it started, makes looking at the individual times less useful. Something may have started at 43s but it doesn’t mean it took 43s to do so… right?
@koying - Lately my zigbee integration has been quite slow to start… and I don’t know why. Still looking into it.
A storm caused my rack mount UPS to fail, and while I was waiting for a warranty replacement (4 days), another storm caused a short power outage right when HA was doing its nightly backup. The next morning there was nothing I could think of that restored HA. I had to abandon the VM and start from scratch which finally allowed me to restore a backup from 2 days prior. In doing so I tried to change things up a bit to improve performance.
Zigbee has been a real pain lately but in this latest snapshot it was surprisingly fast. I think this speed is more of a fluke than the new norm.
I had Chrony as my NTP server and noticed HA waiting >60s for time… I disabled the add-on and installed Chrony on its own VM which appears to have fixed the issue.
I reorganized my 2 proxmox servers so that network services, such as Chrony, Pi-Hole, etc are on one, while applications are on the other trying to build in some more resiliency.
While having a lot of “services” run as an HA addon (such as Chrony) seemed appealing, a major meltdown took everything down so I am reconsidering… individual containers or VMs (to limit the sudden loss and effort needed to restore) seem a safer option.
I need to figure out a better way to backup my VMs… the NAS Proxmox was backing up was down due to the UPS failure and me needing to rewire things temporarily
I am open to any and all suggestions on how to make things safer, better, faster, and hopefully easier to maintain.
