You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@karaf.apache.org by "Xtra Coder (Jira)" <ji...@apache.org> on 2020/01/28 19:33:00 UTC

[jira] [Created] (KARAF-6610) Unstable initial startup of karaf

Xtra Coder created KARAF-6610:
---------------------------------

             Summary: Unstable initial startup of karaf
                 Key: KARAF-6610
                 URL: https://issues.apache.org/jira/browse/KARAF-6610
             Project: Karaf
          Issue Type: Bug
          Components: karaf
         Environment: OS:  CentOS 7.7.1908
Java 11
Karaf 4.2.8 / 4.2.7

My application uses OSGI Declarative Services, ~50 various bundles.

            Reporter: Xtra Coder
         Attachments: 1-startup.txt, 2-manul-restart-of-bundles.txt

We have a problem with unstable initial startup of karaf. From time to time startup fails because of some weird exceptions coming from karaf and felix. Typically after 2nd or at most 3rd restart problem goes away.

I expect there is some race condition depending on CPU speed or something else. That problem almost never occurs on powerful hardware workstations, but is rather frequent on Virtual machines. Because of that issue we have special startup script which retries in case of failures ... and this is somewhat odd.

Today I came across the problem that application under karaf fails to start correctly all the time with any number of retries. After some additional digging i have noticed that there is another process on virtual server which consumes 100% of CPU and therefore ticks left to karaf are very limited. 

To prove idea that karaf failures depend on timings - i killed those processes and karaf started correctly.

After that i've started https://sourceforge.net/projects/systester (./systester-cli -qcborwein 64M -threads 4 -bench) to consume entire CPU and retried starting karaf – it failed.

I’m attaching exceptions I’ve got in the log. There are 2 of them
1.	Initial startup
2.	Admin console displayed all my bundlers as ‘Active’ (strange, but it is) - I restarted few of them manually and got slightly different set of exceptions







--
This message was sent by Atlassian Jira
(v8.3.4#803005)