Bundles not starting waiting forever #4594
Comments
Did you get it to restart at all? When you say "restart" is that an OH restart or a PC/RPi restart? I had a similar thing last week. The system stopped and I restarted OH and many bundles didn't start making the system unusable. I couldn't get OH to restart properly even after a couple of tries, but a complete PC restart fixed it. I'm running on an i86 NUC type box - it's a pretty basic debian install with the absolute minimum running other than OH. I do have logs of my issue - of course there's nothing to say this is the same issue as yours. I did see out of memory exception when this first happened - the system had been running fine with not restarts for around 6 weeks when this happened and wasn't doing anything particularly "stressful" at the time this occurred. After the OH restart I see the following exception -: I didn't clear the cache - I just restarted the PC and it resolved the issue. This exception appears to point to a lock issue - possibly a lockfile (??). I'm not sure why a PC restart would be any different to an OH restart though in this respect. Again - not sure if this is exactly the same issue as @mstormi is seeing, but the symptoms sound similar. |
|
It was only OH restarts. I'm on an RPi 3, but I have seen this on some other Raspis, too, both 32 and 64 openHABian. I have gotten it to fully start after I juggled forth and back with some with startup level definitions. The 'missing reference' message to me indicate it's related to Karaf dependencies and not to file locking, I have not seen any message on that like you did so I'm inclined to believe we're not seeing the same issue. |
I agree in principle, but as I said, I restarted OH twice and that didn't resolve it, so I think I did a soft restart (ie via the OS shutdown command), but I might have done a power cycle - I forget now. Either way, that restored the system.
Ok, useful to know. I guess the bottom line is it's a similar outcome, so I'm interested to hear the answer to your question on how to diagnose. If it happens again I'll dig a bit deeper... |
|
@splatch I read on some occasion in the forum that you know how to translate those missing references. Could you please have a look and explain to us how to decode and to debug this dependency mess? |
|
IIRC the references have the same order as the parameters of the constructor of the service that failed, starting with 0 for the first one. |
|
As it complains about stuff depending on "marketplace", how do I find out which bundle I'm actually from the marketplace and not from openhab-addons package i.e. the co-installed 4.3.2 .kar file? |
|
@mstormi I guess the |
|
I have another instance of the problem. I tried to diag here but don't get anything meaningful out. |
|
how can the marketplace bundle be 'waiting' when it is not installed? |
No output on that, so I think that it was not. Could some core dev in the know please give me guidance how to debug that ? |
|
I noticed Karaf 4.4.7 fixes "avoid race condition at startup between FileInstall and ConfigAdmin" and would love to test if that's related to the issue here. @holgerfriedrich was Karaf 4.4.7 backported into OH 4.3.x? I did not see that in the release notes. Plus I guess backporting is well worth it (regardless of this issue) given we need a stable release as not everybody will happily jump onto OH5 train quick, |
|
@mstormi No, Karaf changes did not get backported. Karaf upgrades touch a lot of dependencies and require quite an amount of testing. It may break plugins compiled before the merge, due to the changing dependencies. For me this is something that goes far beyond the scope of "bugfixes" we typically port back to the last release. Maybe you want to start a discussion about the future of the 4.3.x branch in the light of OH5 breaking changes outside if this issue. Back to this issue: this is something I cannot reproduce. I run two productive 4.3.x instances (w/o using marketplace) and do not encounter the problems described above. I have the core and addons package installed and never clear the caches manually, just let |
Err, no. Patching 4.x up to latest Karaf has nothing to do with "the future" or any breaking changes.
That's not helpful either. OH needs to work on several tens of thousands of instances, and already I myself have more than you that suffer from this problem. And it's not some luxury exotic issue thing. To the best of my knowledge, I don't use anything unusual - actually I am even not aware of using anything from the marketplace - but the issue affects main components like UI and API that everybody needs. Thing is, I have no idea how to diagnose it. |
|
I believe I figured it out. This tarball I've created on a 4.3.2 machine and subsequently been deploying to several freshly installed boxes. When the next oh version got released and the version number shifted up to 4.3.3, openhab 4.3.3 got deployed but these files were extracted from the unchanged tarball and references were thus now pointing to nirvana. I'm sorry for wasting your time as it seems I caused that trouble myself. And yes it's apparently not related to anything-Karaf. |
|
REOPENED. It seems to be an intermittent issue now. After a restart, the system either ends up okay or stuck in this state. |
|
This issue has been mentioned on openHAB Community. There might be relevant details there: https://community.openhab.org/t/openhab-5-0-milestone-discussion/162686/25 |
|
I have actually been dynamically restarting bundles in my code (mostly issued during OH startup), and when I removed that for a test, that 4.3.3 instance I kept having the issue with suddenly went fine. |
I have some bundles that just don't start, resulting in symptoms like an empty homepage (or actually only the UI wizard is offered to run). And I'm seeing this on several OH instances I operate.
It does not happen reproducibly but once it occurs this seems to be a stubborn persistent problem.
Restarts don't help, even if I clean the cache inbetween. Running 4.3.2, I have not seen this in versions earlier than 4.3.
Actually I upgraded one box from Linux pkg 4.3.0-1 (on x86 debian) which did not have the problem to 4.3.2-1 and it showed up.
Any hint how to diagnose what those "missing references $00X" are ?
The text was updated successfully, but these errors were encountered: