-
-
Notifications
You must be signed in to change notification settings - Fork 423
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Snapshot 4069: failed to install bundles at startup #4222
Comments
The problem persists even after a restart. As a reminder, I am using the kar file. Switching back to snapshot 4064 |
@lolodomo I don't think it is a direct consequence, but it certainly does aggrevate the problem. See my analysis here: #4158 (comment) Unless someone figures out how to fix the FeatureInstaller, the only thing that will make a difference is rolling back #4188 (with follow up #4206 and #4209). The install mechanism for finders before was not using the FeatureInstaller. I am not sure this will solve the whole issue though. It looks like it very much appeared wit the upgrade to Karaf 4.4.5. Honestly, this is beyond my knowledge to try to fix. |
Maybe there was something forgotten when moving from Karaf 4.4.4 to 4.4.5 ? |
In the exception, I see "3.18.0.v20220516-2155". Looks like we are referencing version 3.18.0 here: openhab-core/bom/runtime/pom.xml Line 53 in 6eeb9dd
Is the difference important ? I also found this place where version 3.13.100 is defined, I don't know if this is normal.
|
This issue has been mentioned on openHAB Community. There might be relevant details there: |
That looks really strange, why is there a different version here ? @wborn for help |
@lolodomo I don't think that is an issue because we don't run that bundle in a normal installation. @mherwege Continuing here from your #4158 (comment): The problem is that without
The solution was to suppress the bundle refresh after each individual installation and do a refresh after all add-ons have been installed (that should also result in better performance). It seems that since Karaf 4.4.5 there is an issue when the bundles are not immediately refreshed. IMO this is a bug in Karaf. |
@J-N-K @lolodomo If this is a Karaf bug, what can we do to go around it? Would removing the finder installation from the FeatureInstaller again (reverting #4188 with follow up #4206 and #4209) be a remedy, as it would not delay the bundle refresh for finders? Either way, I don't see a good solution at the moment. |
@J-N-K bad news is, Karaf 4.4.6 shows similar behaviour. |
Will the problem persist if the 3 PRs about finder bundles installation are reverted ? |
Since we already saw the issue before (#4158), it might help a bit - but not solve it. We should try to remove the |
#4158 did not seem very critical, as the problem went away by restarting and did not affect all systems (e.g. never seen it on my Windows or RPI systems). With the current changes we have problems which prevent the installation of add-ons in the setup wizard. Just to share some details I learned from the tests I did so far: What I can tell form my tests on Windows is: M2 works for me, and versions up to c316b07 work for me as well (this is last before #4188). 1b503af does not work (#3736). I have not checked the commits between. @J-N-K I tried to remove |
Are you sure the bundle was not reverted to the distribution version? |
I will retry with the clean setup: full distro build, cut off the internet, start openhab. Will take some time... |
Did you also remove Line 202 in 03a9708
The error happens there, and this line should not be required anymore after removing |
this did not do the trick. Did I miss something? |
In my test, I replaced the
It moves the catch up a level and checks for what has been installed and updates the list, even after an error. I didn't expect this to make a difference, but it is worth trying. Apart from that, it is what I did. I didn't try this on the latest snapshot, but on one from 3 days ago, with this patched |
Which build was that? I noticed that this issue didn't exist in 4064, and started in 4065. |
I think it did exist, but it wasn’t as bad as the finders did not get installed at startup due to a bug. Anyway, I am not behind my computer so don’t know the exact version. The only thing I can say is that it did include the changed fixed version of the finder service (which I included manually) and the modified feature installer. It is the bug fix of the finder service that made the problem worse (but did not create it) in the most recent snapshots. And that’s simply because they now install (and didn’t before). But it is the overall feature install that is buggy, probably due to a Karaf version change. |
@mherwege I prepared PRs for rollback of the Karaf upgrade, but it did not help if I remember correctly. I will check again in the evening. |
Yes, confirmed by @J-N-K . I myself see one error at first startup of 4064 but then all addons were finally installed so it was not so critical. |
I just tried to reproduce it with a fresh install (local build of main branches core/distro) and it starts up fine. I then
macOS 14.4.1 / Apple M1 Pro |
I tried it on Mac and it seems to work fine! The problem seems to exist on Linux and Windows. |
@holgerfriedrich @lolodomo Can you comment on OS and Java version? Do you use Oracle, Zulu or Temurin? 64 bit or 32 bit? Arm or x86? |
@J-N-K my "broken" setup is Win11, Zulu 64. I updated to latest, did not make a difference.
|
Can you try temurin? |
Mac + Zulu: works |
Interesting. Can you share the output of Mine looks like that:
|
For me: RPI3, Pi OS, Zulu17, 32 bit |
on Ubuntu 20 + Zulu17 which exbited the problem too:
This is the output of
|
quick status update:
|
@holgerfriedrich Can you try moving l. 198
below the try-catch-block:
In case there are different modifications at the same time, it might be we install new features before the refresh of the old is done. Everthing else looks good IMO. |
I came a bit closer. The difference between both is that the |
Broken install on linux:
Working on mac:
|
@J-N-K confirmed, btw: btw2: |
I tracked it down a bit further. It seems that #3934 is the root cause of the issue. The |
I can confirm the fix. |
Two errors:
This is repeated ever and ever...
The text was updated successfully, but these errors were encountered: