ClassNotFoundException using peer class loading on cluster

classic Classic list List threaded Threaded
2 messages Options
arw180 arw180
Reply | Threaded
Open this post in threaded view
|

ClassNotFoundException using peer class loading on cluster

I'm using peer class loading on a 5 node ignite cluster, persistence enabled, Ignite version 2.8.1. I have a custom class that implements IgniteRunnable and I launch that class on the cluster. This works fine when deploying to an ignite node running on a single node cluster locally, but fails with a ClassNotFound exception (on my custom IgniteRunnable class) on the 5 node cluster. I can see a reference to this class name in both the work-dir/marshaller and work-dir/binary_meta directories on each cluster node, so it seems like the class should be there.

I have many other IgniteRunnables and distributed closures that all work fine -- this is the only one giving me trouble.I tried renaming the class, but that didn't help either.

After nearly three days, I'm running out of ideas (other than giving up and statically deploying the jar to each node, which I really want to avoid), and I'm looking for advice on how to troubleshoot an issue like this.

Thanks for your help,

Alan


vdpyatkov vdpyatkov
Reply | Threaded
Open this post in threaded view
|

Re: ClassNotFoundException using peer class loading on cluster

Hi Alan.

Could you provide a class (IgniteRunnable and may other classes which was laded through p2p with them) which has this unusual behavior?
And please, attache a full stack trace of the exception.

On Thu, Oct 8, 2020 at 5:19 PM Alan Ward <[hidden email]> wrote:
I'm using peer class loading on a 5 node ignite cluster, persistence enabled, Ignite version 2.8.1. I have a custom class that implements IgniteRunnable and I launch that class on the cluster. This works fine when deploying to an ignite node running on a single node cluster locally, but fails with a ClassNotFound exception (on my custom IgniteRunnable class) on the 5 node cluster. I can see a reference to this class name in both the work-dir/marshaller and work-dir/binary_meta directories on each cluster node, so it seems like the class should be there.

I have many other IgniteRunnables and distributed closures that all work fine -- this is the only one giving me trouble.I tried renaming the class, but that didn't help either.

After nearly three days, I'm running out of ideas (other than giving up and statically deploying the jar to each node, which I really want to avoid), and I'm looking for advice on how to troubleshoot an issue like this.

Thanks for your help,

Alan




--
Vladislav Pyatkov