Yes, it’s new in 7.31 and was added to catch mongoose restart and sync with a luup reload. Previously it just stopped accepting connections and required a manual restarting.
Okay and what does that mean a mongoose restart? And what could be the cause?
Mongoose is a service used by the luup engine to store data and when it crashes, the luup engine is restarted.
Okay, and what could be the reason that it crashes? I have enough memory left on my Vera.
I would like to prevent the luup reloads. SInce 2 weeks I have 1 to sometimes 7-8 luup reloads per day. My Vera still functions as it should but before I did not see it Only once in a while.
I seems to recall it’s linked to too many http requests. Back in the alpha stage, I had to remove some code that was doing a lot of http requests to my Vera. I have since removed it, but stability is always high and low on a Vera.
I have to say, I’ve always been a little dubious on this, at least as a generality. I beat the heck out of all my Veras here with a home-spun dashboard, and when I do regression tests on Reactor, I do it with a python script that hammers the system with action and variable requests as fast as it will take them, and I never have seen crashes that I can conclusively relate to all of this traffic. The regression tests can run 30-45 minutes per cycle, flat out and hammering away the whole time. No issues.
I suspect there may be an issue with some particular request, or the way the request is structured/used (i.e. if you’re asking for the entirety of user_data every 30 seconds, you’re gonna have a bad time). If it wasn’t for the fact that I believe at this point we’re unlikely to see a 7.32, I’d say we should dig in more and see if we can isolate it, but at this stage, anything we find has a slim chance of being fixed, I think. At least we’d figure out what to avoid doing, though.
2 Likes
As I reported, in my case it was just updating two variables (color and brightness) for three devices at once. Given the fact that I was running the code with parallelism, and only at night, it was easy for me to catch it.
Moving all the calls from my system to a queue helped a lot. I now have from 2 to 10 days of luup uptime. Unless, out of the blue, it starts to reboot every 20 minutes or a child is corrupted or…
The new fw is too new to be considered at the moment, so we should live with this one, unless we want to move to a different solution. During the last year I wrote so many defensive code to just circumvent odd behaviors…
Yup, we’re all in that boat together. Hopefully we don’t lose the bilge pump.
After puzzling over them for a couple of months, my unexpected Vera restarts now seem to be centered around one particular Reactor routine (called “Lights Out”) that systematically turns off 10+ lights in rapid succession (i.e. no pauses between).
Although the restarts don’t happen every time the routine runs, my theory was that they resulted from the quantity or speed of commands being sent. But then I remembered something…
…my “Lights Out” Reactor routine also included a couple of devices that had been taken offline (unplugged, but not unpaired from Vera) long ago. I had forgotten to remove them, and because Reactor didn’t issue any warnings in this regard, they remained part of the routine (by far, the longest of all my Reactor scenes).
Until today. Following yet another unexpected LUUP restart last night, I went ahead and groomed “Lights Out” by deleting the offline devices. I’ll carefully monitor whether Vera restarts the next few nights, and if she does, I’ll introduce 1-2 second pauses between commands and see if things finally settle.
That may be an aggravation, you’re correct. My “away” routine at night used to turn off all lights (20+) and close all roller shutters (10), unless I divided it into smaller routines. Previously, the chances of a luup reload were very high, after removing the roller shutters it started to behave better. Long story short, dead device may increase the command queue and cause deadlocks, that in Vera’s land equals to a luup reload. As I said, I have this and many more defensive code/actions in place in my controller, but this should not be the case for a stable system… but the current fw generation clearly is not.
1 Like
Mongoose related crashes are starting to happen daily. No changes to the system in several months. The worst part is they are happening right after I set my alarm to away state - so the things I need it to do when I leave home are not completing. I stabilized these routines under UI5 5+ years ago, and then again on UI7 3 years ago.
Nothing to go on in the logs (regular log levels). This time it was
Reload: Downloaded Plugin Critical
…
Quit_set now 1 for 0x15c6750 ThreadedClass::StopThread
…
Mongoose XXX-mg_stop1
…
LuaUPnP Terminated with Exit Code: 0e
Based on this might be time to disable auto update on all my plug-ins. Dang I love/hate Vera.
I disabled it a year ago and never looked back. I usually have from 2 to 10 days of luup uptime, but I had to optimize every single bit of it.
Slightly tangential, but may help
I have (had) a scene that runs at night to turn of all the lights, set the thermostats etc etc. About 15 devices plus some virtual and a couple of TTS.
Triggered by Alexa and another virtual switch or a key press.
Native scene this probably failed to run at all about 20% of the time. Then it would always say ‘not responding’ then it would finish (probably 20 seconds after the ‘not responding’)
I moved it to Reactor yesterday. Finished instantly. Even with additional witty comments via TTS
Might be worth considering replacing Night Mode with a Reactor?
C
1 Like
Chiming in as another user having this issue. @Sorin - I actually have a ticket in with support for the past month or so. Their conclusion was my Nest/WWN plugin (I’m still holding on converting my Nest IDs to Google!).
I have my doubts. I see the Mongoose crashes and mine also appears to be related to Zwave and/or Reactor events (not blaming Reactor - probably something Reactor is telling Vera to do). However, I’ve been unable to correlate my reboots/reloads to any one event.
Frustrating.
I did have a few battery devices that needed to have their batteries replaced (a fob and motion). I did that last night. We’ll see if that helps.
Bumping the tech team. They will be with you shortly. 
3 Likes
“luvd_restart_mongoose” again yesterday. Anyone made any progress on figuring out the reason and how to prevent?
‘Mongoose’ (which I believe is a lightweight web browser built into Linux) has been a problem plaguing Vera users for many years. I seriously doubt any ‘fix’ is in the pipeline.
Here’s a 2016 reference to similar unwanted restarts, which include sa mention of ‘Mongoose’:
I have similar problems and I was promised it will be fixed, but now I think luup engine is basically dead and it has not been in their radar anymore.
1 Like
My ticket has been with senior engineers for a few weeks now. I’m patiently hoping for a fix. I don’t have just reloads (those are almost acceptable) but I’m also having full OS reboots. Argh!
Well patience is a virtue.
I’m going quite the other way
C
1 Like