I'm seeing "lease does not exist" exceptions under some circumstances, e.g.

org.apache.hadoop.ipc.RemoteException: org.apache.hadoop.hbase.regionserver.LeaseException: lease '4341530003498786620' does not exist

After reading some about the exception, I was wondering if the following will cause it. Say I have a very large table with some scattered values I am looking for. I am using a filter to search for these rows. Each call to scanner.next() might need to scan through several million rows before it finds one that matches the filter.

Could this cause the lease to timeout if it takes several minutes to find a filter match, or is there a client-server keep-alive message of some sort that prevents this?

It also looks like this might have been caused by a region server that went down and the cluster didn't recover gracefully.

To follow up, it does appear that the scenario I described will cause the LeaseException, i.e. if you have a large table, and you scan it with a filter that skips many rows, the client can timeout waiting for a response from the region server with the next matching row, which leads to a LeaseException.

I increased hbase.rpc.timeout to resolve the issue.

On Feb 20, 2012, at 4:47 PM, Bryan Keller wrote:

> I'm seeing "lease does not exist" exceptions under some circumstances, e.g.
>
> org.apache.hadoop.ipc.RemoteException: org.apache.hadoop.hbase.regionserver.LeaseException: lease '4341530003498786620' does not exist
>
> After reading some about the exception, I was wondering if the following will cause it. Say I have a very large table with some scattered values I am looking for. I am using a filter to search for these rows. Each call to scanner.next() might need to scan through several million rows before it finds one that matches the filter.
>
> Could this cause the lease to timeout if it takes several minutes to find a filter match, or is there a client-server keep-alive message of some sort that prevents this?
>
> It also looks like this might have been caused by a region server that went down and the cluster didn't recover gracefully.