![]() |
|
Home Assistant addon 1.5.0 crashes soon after startup - Printable Version +- Comfort Automation/ Security System Forums (https://www.comfortforums.com) +-- Forum: Software Applications (https://www.comfortforums.com/forum-4.html) +--- Forum: Comfort with Home Assistant (https://www.comfortforums.com/forum-143.html) +--- Thread: Home Assistant addon 1.5.0 crashes soon after startup (/thread-5656.html) Pages:
1
2
|
Home Assistant addon 1.5.0 crashes soon after startup - Jon798 - 08-03-2025 Hi, System was running as expected with addon 1.4.3. Upgraded to v1.5.0 and now the addon crashes soon after startup. Debug log here: Code: [08:08:08] WARNING: [33mTLS encryption option disabled.[0mNot sure what to try to resolve this? Rolling back to v1.4.3 has it all working again. Thanks in advance for any help. Jon. RE: Home Assistant addon 1.5.0 crashes soon after startup - Ingo - 08-03-2025 You have some sort of connectivity issue to Comfort. Release 1.5.0 has been out for suite some time without any issues. It might just be that due to the connectivity issue the addon eventually crashes. Please make sure you are running the latest Eth03 firmware and have a stable wired network connection. Ingo RE: Home Assistant addon 1.5.0 crashes soon after startup - Jon798 - 08-03-2025 I tried 1.5.0 when it was first released and had the same issue so rolled back to 1.4.3 as I didn't have time to troubleshoot. Tried again today. Running on latest firmware 8.13 and I do have a stable ethernet connection. The v1.5.0 addon runs up fine and then always crashes after startup, it's not like some intermittent connectivity crashes it sometimes, it's every startup and within 30s. Rock solid on v1.4.3, so definitely something in the addon that's different. Debug log doesn't offer any further information that I can use, only the error "cannot access local variable 'original_timeout'..." RE: Home Assistant addon 1.5.0 crashes soon after startup - Ingo - 08-03-2025 There were changes to Keepalives to better report on network stability. I still say there is a network issue that 1.5.0 detects that may have been there since day #1. The following is not normal as Comfort is sending the disconnect. Make sure you don't have anything else connecting to the same port number or try port 1002. 2025-08-03 08:08:09 DEBUG Synchronization Done. 2025-08-03 08:08:19 ERROR Comfort connection error during recv: [Errno 104] Connection reset by peer 2025-08-03 08:08:19 ERROR Comfort Socket Error [Errno 104] Connection reset by peer 2025-08-03 08:08:19 ERROR Lost connection to Comfort, reconnecting... 2025-08-03 08:08:29 INFO Connecting to Comfort (192.168.4.205) on port 1001 2025-08-03 08:08:29 DEBUG LU04 2025-08-03 08:08:29 INFO Comfort Login Ok - User 4 Edit: Please also check your Eth03 idle setting, it might be too short. Code: This software requires a fully functional Comfort Ethernet or Wifi configuration with inactivity timeout set to the default value of 2 minutes. The UCM/Wifi is not recommended due to possible connectivity issues that could arise from switching between different AP's or other possible sources of RF noise. For best performance it is recommended to use either the UCM/Eth03 or the onboard Eth03 Plug-in module on the newer CM9001 Comfort Ultra models. Use a good quality CAT5e or better cable between Comfort and your network device.RE: Home Assistant addon 1.5.0 crashes soon after startup - Jon798 - 08-04-2025 Tried port 1002, no difference. I have a UCM/ETH02 with a wired connection which had timeout set to 0 (disabled). Changed to 2 mins and still same result - addon crashes about 15 seconds after the r?00F00F000000000000000000000000000000000000000000000000000000000000 command is sent (this is 4 bytes shorter than all the other r? commands - assume that's right?) Checked statistics on the switch port the UCM is connected to and the only errors I am seeing are receive packets dropped as the addon crashes. The addon crashes still happen when I don't see the Err104 connection reset by peer message in the logs. Should it report a lost connection rather than crash? Thanks, Jon. RE: Home Assistant addon 1.5.0 crashes soon after startup - Ingo - 08-04-2025 (08-04-2025, 12:52 PM)Jon798 Wrote: Tried port 1002, no difference. Yes, the last r? command is shorter than the others. Packets on Layer-1 should never just be dropped unless there is a line fault or some strange congestion issue where the port cannot handle the traffic and tries to throttle traffic in the only way it knows how, by dropping packets out of the HW queue eg. buffer overruns. This is all in HW so no control on either Comfort or Home Assistant side. With TCP the connection is reset gracefully and all happens very cleanly as you won't see the messages you see when it's a graceful connection drop. The receive packet drops does not sound right. The Err104 is an internal python error code, the Addon tries to send commands in a send routine but at that stage the connection dropped while sending and that is why you see the crash. It does however gracefully shutdown but the actual error is displayed for reference. So to update my previous suspicion, I don't think it's Comfort itself, it might be between Switch port to HA or Switchport to Comfort. The fact that 1.5.0 picks it up is strange, there is nothing is software that makes packet drops show up on a HW port level. Try running wireshark and capture the traffic on that port with the errors, maybe it will give you another clue. Just to be clear, you have an Eth03 and also tested with a Eth02? Both do the same? RE: Home Assistant addon 1.5.0 crashes soon after startup - Jon798 - 08-04-2025 I only have an ETH02, I don't have an ETH03. Analysing wireshark captures is a bit beyond my abilities. I've configured the switch port (Ubiquiti UDM-Pro) to a fixed 10mbit/s full-duplex instead of auto-negotiating in case that helps, and I'll keep an eye on any other errors/drops on the port, but can't shake the feeling that the hardware worked fine with v1.4.3 so ought to work with v1.5.0. Happy to be wrong if I can pinpoint the reason. I guess I could just stay on v1.4.3 but one of the reasons I moved from HomeSeer to Home Assistant was to be on a platform that was keeping pace with changes in other systems, which HomeSeer wasn't. ETH02 only tolerates a single connection at any one time, while ETH03 allows 2 - might that be a factor? Does the addon potentially handle keep alives while zone activations/clearances are being signalled at the same time? Anything else I can try other than blindly trying to upgrade to ETH03 just in case? Thanks, Jon. RE: Home Assistant addon 1.5.0 crashes soon after startup - Ingo - 08-04-2025 No, the Eth03's dual connectivity is totally separate, it doesn't influence the operation. Keepalives and Comfort messages do not interfere at all. The keepalives are sent every 60s (I think) if no command has been sent by HA to Comfort otherwise the connections seems idle to Comfort and it drops it. Edit: I just found an old Eth02 in my box of spares and connected it to my system. I also get the same error and disconnects as you so it seems the Eth02 is truly not fully compatible. Because it's End of Life and End of Support I am not even going to try and find the cause as it might be in the Eth02 firmware which I can't do anything about. My suggestion is to buy a Eth03 module and your issues should be resolved. Edit 2: Did a quick capture and it seems the Eth02 get's hung up at some point. Seems like an internal buffer issue on the Eth02. It just runs out of steam trying to keep up and then gives up. The Addon send routine fails, but keepalives still works for some reason, but Comfort commands get stuck. RE: Home Assistant addon 1.5.0 crashes soon after startup - slychiu - 08-05-2025 r?00F00F000000000000000000000000000000000000000000000000000000000000 This may cause a reset on a non-arm UCM or CM9001. Do you have a new ARM processor? We will be rel;easing a new firmware for CM9001 and UCM RE: Home Assistant addon 1.5.0 crashes soon after startup - Jon798 - 08-05-2025 My CPU and UCM are probably about 20 years old now, so I am guessing the UCM is non-ARM and the CPU is not a 9001. Maybe the potentially troublesome command could be optional? Thanks, Jon. |